From d782dd6fc93b081400e6184b4287e1257a9ebfcc Mon Sep 17 00:00:00 2001 From: Kujtim Hoxha Date: Tue, 28 Oct 2025 17:54:00 +0100 Subject: [PATCH] chore: update test fixtures --- .../anthropic-sonnet/bash_tool.yaml | 72 +-- .../anthropic-sonnet/download_tool.yaml | 110 ++-- .../anthropic-sonnet/glob_tool.yaml | 62 +- .../anthropic-sonnet/grep_tool.yaml | 90 ++- .../anthropic-sonnet/ls_tool.yaml | 81 ++- .../anthropic-sonnet/multiedit_tool.yaml | 194 +++--- .../anthropic-sonnet/parallel_tool_calls.yaml | 89 +-- .../anthropic-sonnet/read_a_file.yaml | 100 +-- .../anthropic-sonnet/simple_test.yaml | 35 +- .../anthropic-sonnet/sourcegraph_tool.yaml | 78 +-- .../anthropic-sonnet/update_a_file.yaml | 193 +++--- .../anthropic-sonnet/write_tool.yaml | 100 +-- .../openai-gpt-5/bash_tool.yaml | 142 +++-- .../openai-gpt-5/download_tool.yaml | 142 ++--- .../openai-gpt-5/glob_tool.yaml | 140 +++-- .../openai-gpt-5/grep_tool.yaml | 136 ++-- .../TestCoderAgent/openai-gpt-5/ls_tool.yaml | 78 +-- .../openai-gpt-5/multiedit_tool.yaml | 528 ++++++++-------- .../openai-gpt-5/parallel_tool_calls.yaml | 174 ++++-- .../openai-gpt-5/read_a_file.yaml | 184 +++--- .../openai-gpt-5/simple_test.yaml | 26 +- .../openai-gpt-5/sourcegraph_tool.yaml | 344 ++++------- .../openai-gpt-5/update_a_file.yaml | 440 +++++++------ .../openai-gpt-5/write_tool.yaml | 136 ++-- .../openrouter-kimi-k2/bash_tool.yaml | 207 ++----- .../openrouter-kimi-k2/download_tool.yaml | 142 +++-- .../openrouter-kimi-k2/glob_tool.yaml | 86 +-- .../openrouter-kimi-k2/grep_tool.yaml | 118 ++-- .../openrouter-kimi-k2/ls_tool.yaml | 100 +-- .../openrouter-kimi-k2/multiedit_tool.yaml | 579 +++++------------- .../parallel_tool_calls.yaml | 148 ++--- .../openrouter-kimi-k2/read_a_file.yaml | 191 +++--- .../openrouter-kimi-k2/simple_test.yaml | 24 +- .../openrouter-kimi-k2/sourcegraph_tool.yaml | 309 ++++------ .../openrouter-kimi-k2/update_a_file.yaml | 300 ++++----- .../openrouter-kimi-k2/write_tool.yaml | 132 ++-- .../TestCoderAgent/zai-glm4.6/bash_tool.yaml | 103 +++- .../zai-glm4.6/download_tool.yaml | 70 +-- .../TestCoderAgent/zai-glm4.6/glob_tool.yaml | 222 +------ .../TestCoderAgent/zai-glm4.6/grep_tool.yaml | 108 ++-- .../TestCoderAgent/zai-glm4.6/ls_tool.yaml | 62 +- .../zai-glm4.6/multiedit_tool.yaml | 324 +++++----- .../zai-glm4.6/parallel_tool_calls.yaml | 210 ++----- .../zai-glm4.6/read_a_file.yaml | 245 +------- .../zai-glm4.6/simple_test.yaml | 64 +- .../zai-glm4.6/sourcegraph_tool.yaml | 146 +++-- .../zai-glm4.6/update_a_file.yaml | 344 ++++------- .../TestCoderAgent/zai-glm4.6/write_tool.yaml | 58 +- 48 files changed, 3492 insertions(+), 4474 deletions(-) diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/bash_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/bash_tool.yaml index afc011346829ca5c2c899742508d0193a0246879..e97a777e881d536d2af19da0eb40bf3f2b33b7e2 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/bash_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/bash_tool.yaml @@ -25,49 +25,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01GMKh3aFc9rXpTdormEcsJZ","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":147,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_012imNSonSLFdbxfQcoTXPvh","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":147,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Bash File"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Bash File"}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Creation"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Creation"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Comman"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Comman"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":147,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":8} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":147,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":8} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 552.794375ms + duration: 507.83925ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45720 + content_length: 44462 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use bash to create a file named test.txt with content 'hello bash'\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use bash to create a file named test.txt with content 'hello bash'\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -84,61 +84,55 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01FwGKwkkgasrRJvxzsZNdxp","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3690,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01MPHcdb5XTN9NdUGpNhthjq","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3690,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01KhTAFFBJTKyXVaaDxyAken","name":"bash","input":{}} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01QbRSeRRHRsGCnViWLy4ywx","name":"bash","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"com"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"mand"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\": \"echo 'h"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"comman"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ello ba"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"d\": \"echo "} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"sh' "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"'hello bash"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"> test"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"' > test.t"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".tx"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"xt\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8757,"output_tokens":60} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8442,"output_tokens":60} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.625279959s + duration: 2.177872625s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46063 + content_length: 44805 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use bash to create a file named test.txt with content 'hello bash'\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01KhTAFFBJTKyXVaaDxyAken\",\"input\":{\"command\":\"echo 'hello bash' \\u003e test.txt\"},\"name\":\"bash\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01KhTAFFBJTKyXVaaDxyAken\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"no output\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use bash to create a file named test.txt with content 'hello bash'\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01QbRSeRRHRsGCnViWLy4ywx\",\"input\":{\"command\":\"echo 'hello bash' \\u003e test.txt\"},\"name\":\"bash\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01QbRSeRRHRsGCnViWLy4ywx\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"no output\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -155,26 +149,26 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_016ghiT5gJ5NFHjDhkHdZAqG","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":72,"cache_read_input_tokens":12447,"cache_creation":{"ephemeral_5m_input_tokens":72,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_011jvZ6Bnn8rrRt2E2SMHjyi","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":72,"cache_read_input_tokens":12132,"cache_creation":{"ephemeral_5m_input_tokens":72,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":72,"cache_read_input_tokens":12447,"output_tokens":4} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":72,"cache_read_input_tokens":12132,"output_tokens":4} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.882943542s + duration: 2.560221792s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/download_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/download_tool.yaml index 229fdc5fa896176faa6e4179d08e3791c6fdf967..df3d850b8844ea75dc126cffe2778bac5772356a 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/download_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/download_tool.yaml @@ -25,34 +25,31 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_0176QErnAZwcxymibhMK4feV","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":160,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01KTVAbrZS28EJmxBd1gj7xp","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":160,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Downloa"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Download example"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d File"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".txt file"}} event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" from"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" from web"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Online Example"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" URL"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" link"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":160,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":9} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":160,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":11} } event: message_stop data: {"type":"message_stop" } @@ -62,15 +59,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 650.846708ms + duration: 885.442833ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45773 + content_length: 44515 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -87,95 +84,74 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01PqWFCfgnjv4C8PxK7D54aD","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3703,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3703,"ephemeral_1h_input_tokens":0},"output_tokens":6,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01B2rT71AcAk7DM8WGBRcWgj","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3703,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3703,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I'll download the file an"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d save it as example.txt."} } - - event: ping - data: {"type": "ping"} - - event: content_block_stop - data: {"type":"content_block_stop","index":0 } - - event: ping - data: {"type": "ping"} - - event: content_block_start - data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_01DHhkBPwN8brz4WnbpoGxwB","name":"download","input":{}} } - - event: ping - data: {"type": "ping"} + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_015URDSHgDU7NFMjnwfxVe3P","name":"download","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""}} event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"url\": "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"url\": \"htt"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"\"h"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ps:"}} event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"ttps://exa"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"//e"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"mple-file"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"xample-"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"s.o"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"files.onlin"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"nl"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"e-convert.c"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"ine-con"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"om/docu"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"vert.c"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ment/txt"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"om/docume"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/example."} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"nt/txt/examp"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"txt\""} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"le.txt\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"file_pa"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":", \"file_p"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"th"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"ath\": \"e"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\": \"exam"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"xa"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ple.tx"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"mple.txt\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t\"}"}} event: content_block_stop - data: {"type":"content_block_stop","index":1 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3703,"cache_read_input_tokens":8757,"output_tokens":104} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3703,"cache_read_input_tokens":8442,"output_tokens":90} } event: message_stop - data: {"type":"message_stop"} + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.3062615s + duration: 3.758014375s - id: 2 request: proto: HTTP/1.1 @@ -221,15 +197,15 @@ interactions: - text/plain; charset=UTF-8 status: 200 OK code: 200 - duration: 188.843959ms + duration: 306.0555ms - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46394 + content_length: 45060 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll download the file and save it as example.txt.\",\"type\":\"text\"},{\"id\":\"toolu_01DHhkBPwN8brz4WnbpoGxwB\",\"input\":{\"file_path\":\"example.txt\",\"url\":\"https://example-files.online-convert.com/document/txt/example.txt\"},\"name\":\"download\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01DHhkBPwN8brz4WnbpoGxwB\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/anthropic-sonnet/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_015URDSHgDU7NFMjnwfxVe3P\",\"input\":{\"file_path\":\"example.txt\",\"url\":\"https://example-files.online-convert.com/document/txt/example.txt\"},\"name\":\"download\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_015URDSHgDU7NFMjnwfxVe3P\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/anthropic-sonnet/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -246,26 +222,26 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01LufXCPJrm7g25PnNRZE3cN","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":4,"cache_creation_input_tokens":161,"cache_read_input_tokens":12460,"cache_creation":{"ephemeral_5m_input_tokens":161,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_014qZcUa2HdyfHS45SPvmAYM","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":4,"cache_creation_input_tokens":148,"cache_read_input_tokens":12145,"cache_creation":{"ephemeral_5m_input_tokens":148,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}}} event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":4,"cache_creation_input_tokens":161,"cache_read_input_tokens":12460,"output_tokens":4} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":4,"cache_creation_input_tokens":148,"cache_read_input_tokens":12145,"output_tokens":4} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.079369708s + duration: 3.107412417s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/glob_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/glob_tool.yaml index 78b0b4831e385240dc7e08fada19b1826f443d94..d997c4a2576ea3307a98fdce703e10f8cf3724cf 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/glob_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/glob_tool.yaml @@ -25,49 +25,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01CPU5CFBx4gU6BfdaNchU9Y","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":142,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_013ZY3WeGSSqGuuJhArn1BGY","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":142,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Finding"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Fin"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Go"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d Go"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Files in Current"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" files in current"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Directory"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" directory"} } event: content_block_stop data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":142,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":9} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":142,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":9} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 550.872667ms + duration: 569.858208ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45709 + content_length: 44451 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files in the current directory\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files in the current directory\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -84,49 +84,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_015k1kGTjzynrDQafHZAoWhc","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3685,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3685,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_015KaLpNT3DBVzPY5bTsCTLX","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3685,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3685,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01PW4ZhRrfq6nED1oPe4z8D9","name":"glob","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_015nhxZ94tKtcrzhqX4SoztW","name":"glob","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"pat"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"pattern"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tern\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\": "} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":": \"*.go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"*"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".go\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3685,"cache_read_input_tokens":8757,"output_tokens":52} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3685,"cache_read_input_tokens":8442,"output_tokens":52} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.11614625s + duration: 2.890654167s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46079 + content_length: 44821 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files in the current directory\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01PW4ZhRrfq6nED1oPe4z8D9\",\"input\":{\"pattern\":\"*.go\"},\"name\":\"glob\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01PW4ZhRrfq6nED1oPe4z8D9\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/glob_tool/main.go\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files in the current directory\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_015nhxZ94tKtcrzhqX4SoztW\",\"input\":{\"pattern\":\"*.go\"},\"name\":\"glob\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_015nhxZ94tKtcrzhqX4SoztW\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/glob_tool/main.go\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -143,16 +143,16 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01K1uo1jgbuP3BTxwm16Cmaa","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":86,"cache_read_input_tokens":12442,"cache_creation":{"ephemeral_5m_input_tokens":86,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01H32VocNk1goX3kkhAduo89","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":86,"cache_read_input_tokens":12127,"cache_creation":{"ephemeral_5m_input_tokens":86,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"main"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d `main.go`"} } event: ping data: {"type": "ping"} @@ -167,14 +167,14 @@ interactions: data: {"type": "ping"} event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":86,"cache_read_input_tokens":12442,"output_tokens":6} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":86,"cache_read_input_tokens":12127,"output_tokens":9} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 6.996937958s + duration: 2.828421916s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/grep_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/grep_tool.yaml index b3b6087f4af8d37bb1eee3cc6c0260b329ad5576..10a272f6968327b0999c3d699aec969307466aab 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/grep_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/grep_tool.yaml @@ -25,52 +25,46 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01TAzqXe9NYNahvhRrX3aTxA","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":144,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01BvK6ycJFfbfr3CZp1Ndtqy","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":144,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Searching "}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Grep"}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"'"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" for 'package'"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"package' in Go Files"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" with"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Grep"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" in Go Files"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":144,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":15} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":144,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":13} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 585.740833ms + duration: 713.37575ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45707 + content_length: 44449 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use grep to search for the word 'package' in go files\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use grep to search for the word 'package' in go files\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -87,58 +81,61 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_011sstAwLRdvnWk6BWCvVD2C","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3687,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3687,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_018UVG6BQMrgaFwmvVN8x5db","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3687,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3687,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01BrQ2xfRkjzhMDSq8KeQtDX","name":"grep","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_019uTzwKcZ5sKbMNCHEY5pGP","name":"grep","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"pattern\":"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"pattern"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" \"p"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\": \"p"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ackage\""}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ack"}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"incl"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"age\""}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ude\": \"*"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"includ"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"e\": \"*.go"}} + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3687,"cache_read_input_tokens":8757,"output_tokens":69} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3687,"cache_read_input_tokens":8442,"output_tokens":69} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop"} headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.3080145s + duration: 2.842227709s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46149 + content_length: 44891 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use grep to search for the word 'package' in go files\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01BrQ2xfRkjzhMDSq8KeQtDX\",\"input\":{\"include\":\"*.go\",\"pattern\":\"package\"},\"name\":\"grep\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01BrQ2xfRkjzhMDSq8KeQtDX\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/anthropic-sonnet/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use grep to search for the word 'package' in go files\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_019uTzwKcZ5sKbMNCHEY5pGP\",\"input\":{\"include\":\"*.go\",\"pattern\":\"package\"},\"name\":\"grep\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_019uTzwKcZ5sKbMNCHEY5pGP\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/anthropic-sonnet/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -155,44 +152,35 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01HTohUehKgMhd4deu5dB7HC","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":125,"cache_read_input_tokens":12444,"cache_creation":{"ephemeral_5m_input_tokens":125,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01JLLgcKi2kWcFexaz1KtfvS","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":125,"cache_read_input_tokens":12129,"cache_creation":{"ephemeral_5m_input_tokens":125,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d 1 match in main.go:"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d 1 match: main"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"1"} } - - event: ping - data: {"type": "ping"} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".go:1"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } - - event: ping - data: {"type": "ping"} - - event: ping - data: {"type": "ping"} + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":125,"cache_read_input_tokens":12444,"output_tokens":14} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":125,"cache_read_input_tokens":12129,"output_tokens":14} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.939075s + duration: 5.08909475s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/ls_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/ls_tool.yaml index 174aa62f59188ed2f60ea60089003db2a84d2af1..4fc828637487416aadd3d095b65a1113b94038b6 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/ls_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/ls_tool.yaml @@ -25,46 +25,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01MxSBvuGVv8LwW2pUXKyHga","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":140,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}}} + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_0126NvBeeJTvRNYzKnSybzB9","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":140,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"List"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"List"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Files"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Files"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" in Current Directory"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" in"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Current Directory"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":140,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":8} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":140,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":8} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop"} headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 603.125375ms + duration: 1.55665475s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45701 + content_length: 44443 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use ls to list the files in the current directory\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use ls to list the files in the current directory\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -81,43 +84,46 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_014oamANtGHWhzgrh6E6maRy","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3683,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3683,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_019NZQJHWQfGYRK2tD91fLPh","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3683,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3683,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01GukFkW1CrLVnWMXbQRxedQ","name":"ls","input":{}}} + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01LXaLR6TwYAPcsN7a2Fn6uq","name":"ls","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"path\":"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" \".\"}"} } + event: ping + data: {"type": "ping"} event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } + + event: ping + data: {"type": "ping"} + + event: ping + data: {"type": "ping"} event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3683,"cache_read_input_tokens":8757,"output_tokens":50} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3683,"cache_read_input_tokens":8442,"output_tokens":34} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.758742125s + duration: 3.031777417s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46085 + content_length: 44817 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use ls to list the files in the current directory\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01GukFkW1CrLVnWMXbQRxedQ\",\"input\":{\"path\":\".\"},\"name\":\"ls\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01GukFkW1CrLVnWMXbQRxedQ\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\n- /tmp/crush-test/TestCoderAgent/anthropic-sonnet/ls_tool/\\n - go.mod\\n - main.go\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use ls to list the files in the current directory\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01LXaLR6TwYAPcsN7a2Fn6uq\",\"input\":{},\"name\":\"ls\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01LXaLR6TwYAPcsN7a2Fn6uq\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\n- /tmp/crush-test/TestCoderAgent/anthropic-sonnet/ls_tool/\\n - go.mod\\n - main.go\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -134,38 +140,29 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01W7AdvpkrUzuBJvhqP1vFMp","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":97,"cache_read_input_tokens":12440,"cache_creation":{"ephemeral_5m_input_tokens":97,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01WgfVTJrbrmnGAtAkpSxUuN","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":81,"cache_read_input_tokens":12125,"cache_creation":{"ephemeral_5m_input_tokens":81,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"go"}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".mod, main.go"} } - - event: ping - data: {"type": "ping"} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".mod, main.go"} } event: content_block_stop - data: {"type":"content_block_stop","index":0} - - event: ping - data: {"type": "ping"} - - event: ping - data: {"type": "ping"} + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":97,"cache_read_input_tokens":12440,"output_tokens":10} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":81,"cache_read_input_tokens":12125,"output_tokens":10} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.854635708s + duration: 2.737991209s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/multiedit_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/multiedit_tool.yaml index 82434040e341ff39e00e13be2b381be519e57260..ada039f5ea4011b258027aac64e345fa9a5a5005 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/multiedit_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/multiedit_tool.yaml @@ -25,52 +25,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01L6UHdS4q4fVV4eePz13WjK","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":170,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_012Aenxf1ob366U6noTThwip","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":170,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } event: content_block_start data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Edit"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Modify"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Go"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Code"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Println"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Modification"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Request"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Greeting"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":170,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":9} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":170,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":10} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 520.542167ms + duration: 1.431345541s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45787 + content_length: 44529 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -87,91 +84,91 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01BGGboUGsf9nf4YB27kr1uD","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3715,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3715,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01MoRw2WbXNv4qef4KsVr12M","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3715,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3715,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I'll rea"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I'll rea"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d the file first to see the exact content an"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d the file first to see the exact content and formatting."} } event: ping data: {"type": "ping"} - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d formatting."} } - event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } + + event: ping + data: {"type": "ping"} event: content_block_start - data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_01Pd1MyYaE6kpBYnPGJ1Fy9p","name":"view","input":{}} } + data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_01RwRCzYE4uwh49CAHjJsViD","name":"view","input":{}} } - event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } + event: ping + data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"file"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"_path\": \"/t"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"file"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"mp/c"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"_path\""} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"rush-test/T"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":": \"/tm"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"estCo"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"p/crush-tes"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"derAgent/an"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"t/TestCode"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"thropi"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"rAgent/a"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"c-sonn"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"nthropic"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"et/multied"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"-sonnet/mult"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"it_tool"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"iedit"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"/mai"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"_tool/main.g"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"n.go\"}"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"o\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":1 } + data: {"type":"content_block_stop","index":1 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3715,"cache_read_input_tokens":8757,"output_tokens":94}} + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3715,"cache_read_input_tokens":8442,"output_tokens":94} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.268274625s + duration: 2.910502166s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46423 + content_length: 45165 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the file first to see the exact content and formatting.\",\"type\":\"text\"},{\"id\":\"toolu_01Pd1MyYaE6kpBYnPGJ1Fy9p\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\"},\"name\":\"view\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01Pd1MyYaE6kpBYnPGJ1Fy9p\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the file first to see the exact content and formatting.\",\"type\":\"text\"},{\"id\":\"toolu_01RwRCzYE4uwh49CAHjJsViD\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\"},\"name\":\"view\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01RwRCzYE4uwh49CAHjJsViD\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -188,142 +185,151 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01G8qnsmV1nmWsKAnBETR9Yd","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":158,"cache_read_input_tokens":12472,"cache_creation":{"ephemeral_5m_input_tokens":158,"ephemeral_1h_input_tokens":0},"output_tokens":45,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01KWJs5wPUmwEj6RizQv8M9W","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":158,"cache_read_input_tokens":12157,"cache_creation":{"ephemeral_5m_input_tokens":158,"ephemeral_1h_input_tokens":0},"output_tokens":28,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_018HUAoLFpcRazziegGrVwai","name":"multiedit","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01Y28kbDbZa4XEu1dibAghuV","name":"multiedit","input":{}} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"file_pa"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"file_pa"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"th\": \"/tmp/c"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"th\": \"/tm"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"rush-t"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"p/cr"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"es"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ush-test/T"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t/TestCo"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"estCoder"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"derA"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Agen"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"gent/"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t/a"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"anth"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"nthropi"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ropic"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"c-sonne"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"-s"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t/multiedi"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"onnet"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t_"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/multiedit_"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tool/"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tool/m"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"main.go\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ain"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \""}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".go\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"edits\": [{"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"edit"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"old_strin"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"s\": [{\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"g\":\"func"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"old_string"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" main() {\\n"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\":"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\\tfmt"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"func mai"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".Println(\\\""}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"n() {\\"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Hell"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"n\\tfmt.Pri"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"o, World!\\\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ntln(\\\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":")\\n}\",\""}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Hello, Wo"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"new"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"rld!\\\")\","}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"_string\":"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"new_"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"f"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"string\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"unc ma"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":":\"func main"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"in() {\\n\\t//"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"() "} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" Gre"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\\n\\t"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"eting\\n\\tfmt"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"// Greet"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".Println(\\"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ing\\n\\tf"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"Hello, "}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"mt."} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Crush!\\\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Print"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":")\\n}\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ln(\\\"Hello, "} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"]}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Cr"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"us"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"h!\\\")\"}]}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":158,"cache_read_input_tokens":12472,"output_tokens":176} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":158,"cache_read_input_tokens":12157,"output_tokens":171} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 4.754166792s + duration: 3.480487083s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 47061 + content_length: 45797 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the file first to see the exact content and formatting.\",\"type\":\"text\"},{\"id\":\"toolu_01Pd1MyYaE6kpBYnPGJ1Fy9p\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\"},\"name\":\"view\",\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01Pd1MyYaE6kpBYnPGJ1Fy9p\",\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_018HUAoLFpcRazziegGrVwai\",\"input\":{\"edits\":[{\"new_string\":\"func main() {\\n\\t// Greeting\\n\\tfmt.Println(\\\"Hello, Crush!\\\")\\n}\",\"old_string\":\"func main() {\\n\\tfmt.Println(\\\"Hello, World!\\\")\\n}\"}],\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\"},\"name\":\"multiedit\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_018HUAoLFpcRazziegGrVwai\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cresult\\u003e\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\\n\\u003c/result\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the file first to see the exact content and formatting.\",\"type\":\"text\"},{\"id\":\"toolu_01RwRCzYE4uwh49CAHjJsViD\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\"},\"name\":\"view\",\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01RwRCzYE4uwh49CAHjJsViD\",\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01Y28kbDbZa4XEu1dibAghuV\",\"input\":{\"edits\":[{\"new_string\":\"func main() {\\n\\t// Greeting\\n\\tfmt.Println(\\\"Hello, Crush!\\\")\",\"old_string\":\"func main() {\\n\\tfmt.Println(\\\"Hello, World!\\\")\"}],\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\"},\"name\":\"multiedit\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01Y28kbDbZa4XEu1dibAghuV\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cresult\\u003e\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool/main.go\\n\\u003c/result\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -340,26 +346,26 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01Pfg51VSKXmk7XDnb5LJxqi","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":221,"cache_read_input_tokens":12630,"cache_creation":{"ephemeral_5m_input_tokens":221,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01Gq7ifzNxxtqk9xDKijiR5n","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":216,"cache_read_input_tokens":12315,"cache_creation":{"ephemeral_5m_input_tokens":216,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":221,"cache_read_input_tokens":12630,"output_tokens":4}} + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":216,"cache_read_input_tokens":12315,"output_tokens":4}} event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.260995042s + duration: 3.858885334s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/parallel_tool_calls.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/parallel_tool_calls.yaml index 0afc67508553cfb469c076f96cda33fd86cfb4e7..9a2a2f560b7d7d0d6d0bcd1de9e963b9ad6fbd55 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/parallel_tool_calls.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/parallel_tool_calls.yaml @@ -25,13 +25,13 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_017pT6NgbTzzXkCGSTQshTPp","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":159,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_0134PDJ3fpTMLgTkhFfsFouJ","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":159,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Parallel"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Parallel"} } event: content_block_delta data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Go File"} } @@ -40,34 +40,37 @@ interactions: data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Discovery"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Listing with"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" and Listing"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Glob and "} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"LS"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0} event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":159,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":11} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":159,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":15} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 563.628166ms + duration: 598.373334ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45798 + content_length: 44540 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -84,61 +87,64 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_0194xppZMNBYpR4cTqu2EVf9","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3704,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01BDfizVzHgt1Hkjbk3vy5cK","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3704,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01RXPG7ekJ6r23yMpNyaWZ1T","name":"glob","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01MMZ8TvGbhRxEatX2YCguxp","name":"glob","input":{}} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"pa"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"pattern\": "} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ttern\": "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"**/*."} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"**/*.go\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"go\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: content_block_start - data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_019gc7awTgkJ8XcBF7L982ug","name":"ls","input":{}} } + data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_011VRmbQhhy9VxKfuk8zEP9C","name":"ls","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"pat"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"p"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"h\": \".\"}"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"ath\": \".\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":1 } + data: {"type":"content_block_stop","index":1 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8757,"output_tokens":86} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8442,"output_tokens":86} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.244434834s + duration: 2.184773291s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46479 + content_length: 45221 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01RXPG7ekJ6r23yMpNyaWZ1T\",\"input\":{\"pattern\":\"**/*.go\"},\"name\":\"glob\",\"type\":\"tool_use\"},{\"id\":\"toolu_019gc7awTgkJ8XcBF7L982ug\",\"input\":{\"path\":\".\"},\"name\":\"ls\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01RXPG7ekJ6r23yMpNyaWZ1T\",\"content\":[{\"text\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls/main.go\",\"type\":\"text\"}],\"type\":\"tool_result\"},{\"tool_use_id\":\"toolu_019gc7awTgkJ8XcBF7L982ug\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\n- /tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01MMZ8TvGbhRxEatX2YCguxp\",\"input\":{\"pattern\":\"**/*.go\"},\"name\":\"glob\",\"type\":\"tool_use\"},{\"id\":\"toolu_011VRmbQhhy9VxKfuk8zEP9C\",\"input\":{\"path\":\".\"},\"name\":\"ls\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01MMZ8TvGbhRxEatX2YCguxp\",\"content\":[{\"text\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls/main.go\",\"type\":\"text\"}],\"type\":\"tool_result\"},{\"tool_use_id\":\"toolu_011VRmbQhhy9VxKfuk8zEP9C\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\n- /tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -155,35 +161,44 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01Q7JbYpzEHTGw2DvPp1LDSQ","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":6,"cache_creation_input_tokens":207,"cache_read_input_tokens":12461,"cache_creation":{"ephemeral_5m_input_tokens":207,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01QaQ34fH2Pkjzk5r1gGWVtm","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":6,"cache_creation_input_tokens":207,"cache_read_input_tokens":12146,"cache_creation":{"ephemeral_5m_input_tokens":207,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d one Go file: `main.go`. Directory"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d one Go file (main.go) in the directory, which also contains"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" contains `go.mod` and `main.go`."} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" go.mod."} } + + event: ping + data: {"type": "ping"} event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } + + event: ping + data: {"type": "ping"} + + event: ping + data: {"type": "ping"} event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":6,"cache_creation_input_tokens":207,"cache_read_input_tokens":12461,"output_tokens":26} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":6,"cache_creation_input_tokens":207,"cache_read_input_tokens":12146,"output_tokens":23} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.10474175s + duration: 2.5900015s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/read_a_file.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/read_a_file.yaml index ade7ff4bc85fc86ad34b0ab53fd5edca359eb78b..4241b11578dafd50d35d009df04d6ec8ca587c99 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/read_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/read_a_file.yaml @@ -25,46 +25,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01FPxG45ganMpU7HdKFFJRAX","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":134,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01EfEc4NSjxHgVL5Rx5tJKCS","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":134,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Rea"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Rea"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d Go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d Go"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Mod Requirements"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Module"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Details"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":134,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":8} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":134,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":7} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 675.620792ms + duration: 914.98375ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45671 + content_length: 44413 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"Read the go mod\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"Read the go mod\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -81,79 +84,73 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_011vGHdc49fZFTncSji6gfpZ","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3679,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3679,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01JNfkrR5zfAjKPU7tYz91e1","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3679,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3679,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_0127pa1AoNFjBcwBw4zYLKja","name":"view","input":{}} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01B3gNMN9RDniJMb735cD6rG","name":"view","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"file"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"_path\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"file_pat"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":": "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"h\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"/tmp"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":": \"/tmp/c"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/crush-te"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"rush-te"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"st/Test"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"st/TestC"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Co"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"oderA"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"derA"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ge"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"gent/an"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"nt/anthro"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"thropic-son"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"pic-sonn"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"net/read"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"et/read_a_"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"_a_fi"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"file/go.mod"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"le/g"}} - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"o.mod\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0} + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3679,"cache_read_input_tokens":8757,"output_tokens":79} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3679,"cache_read_input_tokens":8442,"output_tokens":79} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.347076542s + duration: 2.588140125s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46141 + content_length: 44883 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"Read the go mod\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_0127pa1AoNFjBcwBw4zYLKja\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/read_a_file/go.mod\"},\"name\":\"view\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_0127pa1AoNFjBcwBw4zYLKja\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"Read the go mod\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01B3gNMN9RDniJMb735cD6rG\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/read_a_file/go.mod\"},\"name\":\"view\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01B3gNMN9RDniJMb735cD6rG\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -170,35 +167,44 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_014d1eZhc3W5RKoHAvZuT7EP","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":120,"cache_read_input_tokens":12436,"cache_creation":{"ephemeral_5m_input_tokens":120,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_015PrPiKEYe6AtwwRjEWdeQ8","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":120,"cache_read_input_tokens":12121,"cache_creation":{"ephemeral_5m_input_tokens":120,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Module"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Module"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" is `example.com/testproject` on"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":": `example.com/testproject`,"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Go 1.23, no dependencies."} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Go 1.23"} } + + event: ping + data: {"type": "ping"} event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } + + event: ping + data: {"type": "ping"} + + event: ping + data: {"type": "ping"} event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":120,"cache_read_input_tokens":12436,"output_tokens":23} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":120,"cache_read_input_tokens":12121,"output_tokens":18} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 5.744573083s + duration: 4.032176s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/simple_test.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/simple_test.yaml index ad2bdc2c2d582c840fd403b0c6ce394991a4d13e..811e87cd0042cb0dece3b66a8aa1131f4e66a669 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/simple_test.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/simple_test.yaml @@ -25,22 +25,22 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01FA7ZZHsnXN3riuJK21jkFB","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":131,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}}} + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01FPjA3RpMzRGubxYbwbFP1N","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":131,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Quick"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Quick"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Hello"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Hello"} } event: ping data: {"type": "ping"} event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: ping data: {"type": "ping"} @@ -52,22 +52,22 @@ interactions: data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":131,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":5} } event: message_stop - data: {"type":"message_stop"} + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 567.335625ms + duration: 1.261823625s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45661 + content_length: 44403 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"Hello\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"Hello\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -84,29 +84,26 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01GzNjpZKuLqwVRuTXLJWD9r","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3674,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3674,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01GghZV67GWJCjHUCU2Jqkhu","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":12116,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":12116,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Hello! How"} } - - event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" can I help you today?"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Hello"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3674,"cache_read_input_tokens":8757,"output_tokens":12} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":12116,"cache_read_input_tokens":0,"output_tokens":4} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.373099208s + duration: 3.402157917s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/sourcegraph_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/sourcegraph_tool.yaml index 44949ada65ec7db04853a1d8a95a6df86917e473..7c898be0365361d7d29d68211326d6870f7c3252 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/sourcegraph_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/sourcegraph_tool.yaml @@ -25,49 +25,49 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01DW7x9Bihrph6LzBucGdfeE","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01PaXxHh4BqrWiBMGLqxJw5p","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":5,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Search Go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Sourcegraph Go"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Repos"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Repo"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" for Main"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Main"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Functions"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Function Search"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":10} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":13} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 537.344291ms + duration: 886.270708ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45721 + content_length: 44463 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use sourcegraph to search for 'func main' in Go repositories\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use sourcegraph to search for 'func main' in Go repositories\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -84,37 +84,37 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01GxPbWwzaww3bcLnXysTZi7","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3690,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01W6bwfuWeKiBXBP7FkDht7n","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3690,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01Fgv2C8Gdoym6CxBDEUR2FH","name":"sourcegraph","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01TqXhpQymyHsJ4dYQgjavjz","name":"sourcegraph","input":{}} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"query\": \""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"query\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"lang:go fun"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":": \"lan"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"c mai"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"g:go func ma"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"n\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"in\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"count\": 1"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"cou"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"0}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"nt\": 10}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8757,"output_tokens":74} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8442,"output_tokens":74} } event: message_stop data: {"type":"message_stop" } @@ -124,7 +124,7 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 4.503319375s + duration: 3.035625833s - id: 2 request: proto: HTTP/1.1 @@ -152,15 +152,15 @@ interactions: - application/json status: 200 OK code: 200 - duration: 1.429988s + duration: 1.485277292s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 53517 + content_length: 52259 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use sourcegraph to search for 'func main' in Go repositories\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01Fgv2C8Gdoym6CxBDEUR2FH\",\"input\":{\"count\":10,\"query\":\"lang:go func main\"},\"name\":\"sourcegraph\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01Fgv2C8Gdoym6CxBDEUR2FH\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| func main() {\\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func greeting() {\\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n## Result 2: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/02_func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 3: github.com/jcalabro/uscope/assets/gobacktrace/main.go\\n\\nURL: /github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go\\n\\n```\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| func main() {\\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| \\tFuncA()\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncB()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncC()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncD()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncE()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncF()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func FuncE() {\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\tfmt.Println(\\\"FuncE\\\")\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| func FuncD() {\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| \\tFuncE()\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tfmt.Println(\\\"FuncD\\\")\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n```\\n\\n```\\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| func FuncC() {\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n```\\n\\n```\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| \\tFuncD()\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n```\\n\\n```\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tfmt.Println(\\\"FuncC\\\")\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n```\\n\\n```\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| func FuncB() {\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n```\\n\\n```\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| \\tFuncC()\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n```\\n\\n```\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tfmt.Println(\\\"FuncB\\\")\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n```\\n\\n```\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| func FuncA() {\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n```\\n\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use sourcegraph to search for 'func main' in Go repositories\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01TqXhpQymyHsJ4dYQgjavjz\",\"input\":{\"count\":10,\"query\":\"lang:go func main\"},\"name\":\"sourcegraph\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01TqXhpQymyHsJ4dYQgjavjz\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| func main() {\\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func greeting() {\\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n## Result 2: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/02_func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 3: github.com/jcalabro/uscope/assets/gobacktrace/main.go\\n\\nURL: /github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go\\n\\n```\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| func main() {\\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| \\tFuncA()\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncB()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncC()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncD()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncE()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncF()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func FuncE() {\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\tfmt.Println(\\\"FuncE\\\")\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| func FuncD() {\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| \\tFuncE()\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tfmt.Println(\\\"FuncD\\\")\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n```\\n\\n```\\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| func FuncC() {\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n```\\n\\n```\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| \\tFuncD()\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n```\\n\\n```\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tfmt.Println(\\\"FuncC\\\")\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n```\\n\\n```\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| func FuncB() {\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n```\\n\\n```\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| \\tFuncC()\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n```\\n\\n```\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tfmt.Println(\\\"FuncB\\\")\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n```\\n\\n```\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| func FuncA() {\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n```\\n\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -177,44 +177,44 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_014724o9o3akbVQp5S8V2bXU","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":3658,"cache_read_input_tokens":12447,"cache_creation":{"ephemeral_5m_input_tokens":3658,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_019bCHRCAFq86GqmgPF4aw9E","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":3658,"cache_read_input_tokens":12132,"cache_creation":{"ephemeral_5m_input_tokens":3658,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Foun"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d 30 matches for"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d 30 matches across Go"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" `func main` in Go repositories. Results"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" repositories. Common patterns include simple \"Hello"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" include training examples from GolangTraining repository"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" World\" programs, function composition examples"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" showing basic main function usage an"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":", and test programs with multiple function calls from"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d a backtrace testing program."} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" main."} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":3658,"cache_read_input_tokens":12447,"output_tokens":41} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":3658,"cache_read_input_tokens":12132,"output_tokens":36} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.144470791s + duration: 9.121800958s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/update_a_file.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/update_a_file.yaml index 1195710a582adca3168b340107b1092f7729f24a..e234933377dd4d4051190d09d272a1db81c2bb5d 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/update_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/update_a_file.yaml @@ -25,52 +25,55 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01W5jKwUPQD2FLMK2WSdGvHR","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01JX9hGcaLZw6JEJeqvLK2D9","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":2,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Update"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Modify"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Hello"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" main"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Print"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".go to print"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" in"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Hello"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" main.go"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" from"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" Crush"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":10} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":145,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":14} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 674.908625ms + duration: 549.701125ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45727 + content_length: 44469 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"update the main.go file by changing the print to say hello from crush\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"update the main.go file by changing the print to say hello from crush\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -87,94 +90,97 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01B3DXnXZibVscXxLJhzYtgg","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3690,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01MqSPwH6d8D6C4vNna59Bgm","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3690,"ephemeral_1h_input_tokens":0},"output_tokens":3,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I'll rea"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I'll rea"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d the file first to see its current content, then update it."} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"d the main.go file first to see"} } event: ping data: {"type": "ping"} - event: content_block_stop - data: {"type":"content_block_stop","index":0 } + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" its current content, then update it."} } - event: ping - data: {"type": "ping"} + event: content_block_stop + data: {"type":"content_block_stop","index":0 } event: content_block_start - data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_01SivyggTSfK7BShmdYXZnip","name":"view","input":{}} } + data: {"type":"content_block_start","index":1,"content_block":{"type":"tool_use","id":"toolu_01KN4SjwjijmUS51AFEgR9hc","name":"view","input":{}} } - event: ping - data: {"type": "ping"} + event: content_block_delta + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } + + event: content_block_delta + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"file_"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":""} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"path\": \"/tm"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"{\"fi"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"p/crush-te"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"le_pat"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"st/Tes"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"h\": \"/t"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"tCoderA"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"mp/crush-"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"gent/"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"test/TestCod"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"anth"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"erA"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"ropic-"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"ge"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"so"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"nt/anthropi"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"nnet/"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"c-so"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"updat"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"nnet/update_"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"e_a_file"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"a_file"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"/main.g"} } event: content_block_delta - data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"/main.go\"}"} } + data: {"type":"content_block_delta","index":1,"delta":{"type":"input_json_delta","partial_json":"o\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":1 } + data: {"type":"content_block_stop","index":1 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8757,"output_tokens":96}} + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3690,"cache_read_input_tokens":8442,"output_tokens":99} } event: message_stop - data: {"type":"message_stop"} + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.252981s + duration: 3.407652375s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46365 + content_length: 45115 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"update the main.go file by changing the print to say hello from crush\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the file first to see its current content, then update it.\",\"type\":\"text\"},{\"id\":\"toolu_01SivyggTSfK7BShmdYXZnip\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\"},\"name\":\"view\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01SivyggTSfK7BShmdYXZnip\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"update the main.go file by changing the print to say hello from crush\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the main.go file first to see its current content, then update it.\",\"type\":\"text\"},{\"id\":\"toolu_01KN4SjwjijmUS51AFEgR9hc\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\"},\"name\":\"view\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01KN4SjwjijmUS51AFEgR9hc\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -191,130 +197,139 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01FrFSHBSdc8UAwrfb1knGDK","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":160,"cache_read_input_tokens":12447,"cache_creation":{"ephemeral_5m_input_tokens":160,"ephemeral_1h_input_tokens":0},"output_tokens":26,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_014ueD6MywmVnsrM5ma23M1K","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":163,"cache_read_input_tokens":12132,"cache_creation":{"ephemeral_5m_input_tokens":163,"ephemeral_1h_input_tokens":0},"output_tokens":33,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01Y6UdvSDasvmGwtK5A5KNTe","name":"edit","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_011HkyGjpcSAk9bMHg9DvGfU","name":"edit","input":{}} } event: content_block_delta data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"file_path\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"f"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ile_"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"path\": \"/"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tmp"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":": "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/crush-test"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"/tm"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/TestCoderAg"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"p/cru"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ent/a"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"sh-test/Test"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"nt"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"CoderAgent"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"hropic-s"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/anthrop"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"onnet/"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ic-s"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"update_a"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"onn"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"_fil"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"et/update_a"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"e/main.go\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"_file/main"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":".go\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"old_string\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"o"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":": \"func main"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ld_string\":"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"() {\\n\\tfmt."} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" \"func "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Pr"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"main() {\\n"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"int"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\\tfmt.Prin"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ln"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tln(\\\"Hello,"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"(\\\"He"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" Wor"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"llo, Worl"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ld!\\\")\\n"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"d!\\\")\\n}\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"}\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"new"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"new_st"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"_string\":"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ring\": \"f"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" \"func m"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"unc"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ain() {\\n\\tf"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" main() "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"mt.Pri"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\\n\\tfmt.Pri"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"nt"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ntln("} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ln("} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\\\"hel"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\\\"hello"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"lo from "} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" fr"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"crush\\"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"om crush"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\")\\n}\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\\\")\\n}\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0} event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":160,"cache_read_input_tokens":12447,"output_tokens":147} } + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":163,"cache_read_input_tokens":12132,"output_tokens":147} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.474816458s + duration: 2.933049583s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46973 + content_length: 45723 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"update the main.go file by changing the print to say hello from crush\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the file first to see its current content, then update it.\",\"type\":\"text\"},{\"id\":\"toolu_01SivyggTSfK7BShmdYXZnip\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\"},\"name\":\"view\",\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01SivyggTSfK7BShmdYXZnip\",\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01Y6UdvSDasvmGwtK5A5KNTe\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\",\"new_string\":\"func main() {\\n\\tfmt.Println(\\\"hello from crush\\\")\\n}\",\"old_string\":\"func main() {\\n\\tfmt.Println(\\\"Hello, World!\\\")\\n}\"},\"name\":\"edit\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01Y6UdvSDasvmGwtK5A5KNTe\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cresult\\u003e\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\\n\\u003c/result\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"update the main.go file by changing the print to say hello from crush\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"text\":\"I'll read the main.go file first to see its current content, then update it.\",\"type\":\"text\"},{\"id\":\"toolu_01KN4SjwjijmUS51AFEgR9hc\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\"},\"name\":\"view\",\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01KN4SjwjijmUS51AFEgR9hc\",\"content\":[{\"text\":\"\\u003cfile\\u003e\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\u003c/file\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_011HkyGjpcSAk9bMHg9DvGfU\",\"input\":{\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\",\"new_string\":\"func main() {\\n\\tfmt.Println(\\\"hello from crush\\\")\\n}\",\"old_string\":\"func main() {\\n\\tfmt.Println(\\\"Hello, World!\\\")\\n}\"},\"name\":\"edit\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_011HkyGjpcSAk9bMHg9DvGfU\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cresult\\u003e\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file/main.go\\n\\u003c/result\\u003e\\n\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -331,26 +346,26 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01T8L3Lmwz5cHpp4sbtcZA1Z","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":199,"cache_read_input_tokens":12607,"cache_creation":{"ephemeral_5m_input_tokens":199,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01NJhtm9P7WZfZRonXEiR7zj","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":5,"cache_creation_input_tokens":199,"cache_read_input_tokens":12295,"cache_creation":{"ephemeral_5m_input_tokens":199,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}} + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":199,"cache_read_input_tokens":12607,"output_tokens":4} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":5,"cache_creation_input_tokens":199,"cache_read_input_tokens":12295,"output_tokens":4} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.310453041s + duration: 3.03882125s diff --git a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/write_tool.yaml b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/write_tool.yaml index ac5714c0d4a0392a4ffed6268529b6960b469fa8..31cb8508696bc4bc6f4f8bc3a945a7893b984bdf 100644 --- a/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/write_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/anthropic-sonnet/write_tool.yaml @@ -25,49 +25,52 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01EygE1KGZz3X3ibj5Yf4RLN","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":161,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-3-5-haiku-20241022","id":"msg_01VVH7WPqzJXcfSxCuksGTUm","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":161,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"cache_creation":{"ephemeral_5m_input_tokens":0,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Create"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Creating"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" config"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" config"} } event: ping data: {"type": "ping"} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".json with JSON"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":".json with"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" data"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" JSON"} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" content"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":161,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":10} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":161,"cache_creation_input_tokens":0,"cache_read_input_tokens":0,"output_tokens":10} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 574.209875ms + duration: 655.177042ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45764 + content_length: 44506 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -84,97 +87,100 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01U58TsX9NNkQuubNnpYnqqf","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8757,"cache_creation":{"ephemeral_5m_input_tokens":3704,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_018UjgTTetHdN1STit86Z4MV","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8442,"cache_creation":{"ephemeral_5m_input_tokens":3704,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_015v3pfbXeD2N7WUz1d5Lkyj","name":"write","input":{}} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"tool_use","id":"toolu_01GNpsQYC1xJAY4wqRHKqiSv","name":"write","input":{}} } event: content_block_delta data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\"file_pat"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"{\""} } + + event: content_block_delta + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"file_p"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"h\": \"/tmp"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ath\": "} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/crush-test/"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"/t"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"Te"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"mp/c"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"stCoderAgent"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"rush-te"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/anthro"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"st/Tes"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"pic-sonne"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tCoderAgen"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t/write_too"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"t/anthr"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"l/con"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"opic-"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"fig.json\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"sonnet"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"conte"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"/write_tool/"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"nt\": \"{\\\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"config.json\""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"name\\\":"}} + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":", \"con"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" \\"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"tent\": \"{\\\"n"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"test\\\", \\\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ame\\"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"ve"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\": \\\"test\\"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"rsion\\\": \\"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\", \\\"versio"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"1.0."} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"n\\\":"} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"0\\\""} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":" \\\"1.0.0\\\"}"}} event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"}\"}"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"input_json_delta","partial_json":"\"}"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8757,"output_tokens":109}} + data: {"type":"message_delta","delta":{"stop_reason":"tool_use","stop_sequence":null},"usage":{"input_tokens":2,"cache_creation_input_tokens":3704,"cache_read_input_tokens":8442,"output_tokens":109} } event: message_stop - data: {"type":"message_stop"} + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.586460958s + duration: 2.934821792s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46333 + content_length: 45075 host: "" - body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_015v3pfbXeD2N7WUz1d5Lkyj\",\"input\":{\"content\":\"{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}\",\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool/config.json\"},\"name\":\"write\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_015v3pfbXeD2N7WUz1d5Lkyj\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cresult\\u003e\\nFile successfully written: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool/config.json\\n\\u003c/result\\u003e\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"},\"name\":\"web_fetch\",\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\u003cusage\\u003e\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (\\u003e50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" + body: "{\"max_tokens\":10000,\"messages\":[{\"content\":[{\"text\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"type\":\"text\"}],\"role\":\"user\"},{\"content\":[{\"id\":\"toolu_01GNpsQYC1xJAY4wqRHKqiSv\",\"input\":{\"content\":\"{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}\",\"file_path\":\"/tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool/config.json\"},\"name\":\"write\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"tool_use\"}],\"role\":\"assistant\"},{\"content\":[{\"tool_use_id\":\"toolu_01GNpsQYC1xJAY4wqRHKqiSv\",\"cache_control\":{\"type\":\"ephemeral\"},\"content\":[{\"text\":\"\\u003cresult\\u003e\\nFile successfully written: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool/config.json\\n\\u003c/result\\u003e\",\"type\":\"text\"}],\"type\":\"tool_result\"}],\"role\":\"user\"}],\"model\":\"claude-sonnet-4-5-20250929\",\"system\":[{\"text\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\u003ccritical_rules\\u003e\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\u003c/critical_rules\\u003e\\n\\n\\u003ccommunication_style\\u003e\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\u003c/communication_style\\u003e\\n\\n\\u003ccode_references\\u003e\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\u003c/code_references\\u003e\\n\\n\\u003cworkflow\\u003e\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\u003c/workflow\\u003e\\n\\n\\u003cdecision_making\\u003e\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\u003c/decision_making\\u003e\\n\\n\\u003ctask_scope\\u003e\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\u003c/task_scope\\u003e\\n\\n\\u003cediting_files\\u003e\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\u003c/editing_files\\u003e\\n\\n\\u003cwhitespace_and_exact_matching\\u003e\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\u003c/whitespace_and_exact_matching\\u003e\\n\\n\\u003cerror_handling\\u003e\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\u003c/error_handling\\u003e\\n\\n\\u003cmemory_instructions\\u003e\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\u003c/memory_instructions\\u003e\\n\\n\\u003ccode_conventions\\u003e\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\u003c/code_conventions\\u003e\\n\\n\\u003ctesting\\u003e\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\u003c/testing\\u003e\\n\\n\\u003ctool_usage\\u003e\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\u003cbash_commands\\u003e\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `\\u0026` for background processes that won't stop on their own (e.g., `node server.js \\u0026`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status \\u0026\\u0026 git diff HEAD \\u0026\\u0026 git log -n 3`)\\n\\u003c/bash_commands\\u003e\\n\\u003c/tool_usage\\u003e\\n\\n\\u003cproactiveness\\u003e\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\u003c/proactiveness\\u003e\\n\\n\\u003cfinal_answers\\u003e\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\u003c/final_answers\\u003e\\n\\n\\u003cenv\\u003e\\nWorking directory: /tmp/crush-test/TestCoderAgent/anthropic-sonnet/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\u003c/env\\u003e\\n\\n\\n\\n\\n\",\"cache_control\":{\"type\":\"ephemeral\"},\"type\":\"text\"}],\"tool_choice\":{\"disable_parallel_tool_use\":false,\"type\":\"auto\"},\"tools\":[{\"input_schema\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"},\"name\":\"bash\",\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\u003ccross_platform\\u003e\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\u003c/cross_platform\\u003e\\n\\n\\u003cexecution_steps\\u003e\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with \\u003ccwd\\u003e\\u003c/cwd\\u003e tags\\n\\u003c/execution_steps\\u003e\\n\\n\\u003cusage_notes\\u003e\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '\\u0026\\u0026', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\u003c/usage_notes\\u003e\\n\\n\\u003cgit_commits\\u003e\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in \\u003ccommit_analysis\\u003e tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\u003ccrush@charm.land\\u003e\\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\u003c/git_commits\\u003e\\n\\n\\u003cpull_requests\\u003e\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in \\u003cpr_analysis\\u003e tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat \\u0026lt;\\u0026lt;'EOF'\\n\\n ## Summary\\n\\n \\u0026lt;1-3 bullet points\\u003e\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\u003c/pull_requests\\u003e\\n\\n\\u003cexamples\\u003e\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar \\u0026\\u0026 pytest tests\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"},\"name\":\"download\",\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\u003cusage\\u003e\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"},\"name\":\"edit\",\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\u003c/parameters\\u003e\\n\\n\\u003cspecial_cases\\u003e\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\u003c/special_cases\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\u003c/critical_requirements\\u003e\\n\\n\\u003cwarnings\\u003e\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) \\u003e 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\u003c/examples\\u003e\\n\\n\\u003cwindows_notes\\u003e\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\u003c/windows_notes\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"},\"name\":\"multiedit\",\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\u003cprerequisites\\u003e\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\u003c/prerequisites\\u003e\\n\\n\\u003cparameters\\u003e\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\u003c/parameters\\u003e\\n\\n\\u003coperation\\u003e\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\u003c/operation\\u003e\\n\\n\\u003cinherited_rules\\u003e\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\u003c/inherited_rules\\u003e\\n\\n\\u003ccritical_requirements\\u003e\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\u003c/critical_requirements\\u003e\\n\\n\\u003cverification_before_using\\u003e\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\u003c/verification_before_using\\u003e\\n\\n\\u003cwarnings\\u003e\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\u003c/warnings\\u003e\\n\\n\\u003crecovery_steps\\u003e\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\u003c/recovery_steps\\u003e\\n\\n\\u003cbest_practices\\u003e\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\u003c/best_practices\\u003e\\n\\n\\u003cwhitespace_checklist\\u003e\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\u003c/whitespace_checklist\\u003e\\n\\n\\u003cexamples\\u003e\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\u003c/examples\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"glob\",\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cpattern_syntax\\u003e\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\u003c/pattern_syntax\\u003e\\n\\n\\u003cexamples\\u003e\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\u003c/examples\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"},\"name\":\"grep\",\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\u003cusage\\u003e\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\u003c/usage\\u003e\\n\\n\\u003cregex_syntax\\u003e\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\u003c/regex_syntax\\u003e\\n\\n\\u003cinclude_patterns\\u003e\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\u003c/include_patterns\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\u003c/limitations\\u003e\\n\\n\\u003cignore_support\\u003e\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\u003c/ignore_support\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"},\"name\":\"ls\",\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\u003cusage\\u003e\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"},\"name\":\"sourcegraph\",\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\u003cusage\\u003e\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\u003c/usage\\u003e\\n\\n\\u003cbasic_syntax\\u003e\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\u003c/basic_syntax\\u003e\\n\\n\\u003ckey_filters\\u003e\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\u003c/key_filters\\u003e\\n\\n\\u003cexamples\\u003e\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\u003c/examples\\u003e\\n\\n\\u003cboolean_operators\\u003e\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\u003c/boolean_operators\\u003e\\n\\n\\u003climitations\\u003e\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\u003c/limitations\\u003e\\n\\n\\u003ctips\\u003e\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"},\"name\":\"view\",\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\u003cusage\\u003e\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines \\u003e2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\u003c/tips\\u003e\\n\"},{\"input_schema\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"},\"name\":\"write\",\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\u003cusage\\u003e\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\u003c/usage\\u003e\\n\\n\\u003cfeatures\\u003e\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\u003c/features\\u003e\\n\\n\\u003climitations\\u003e\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\u003c/limitations\\u003e\\n\\n\\u003ccross_platform\\u003e\\n- Use forward slashes (/) for compatibility\\n\\u003c/cross_platform\\u003e\\n\\n\\u003ctips\\u003e\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\u003c/tips\\u003e\\n\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"stream\":true}" headers: Accept: - application/json @@ -191,26 +197,26 @@ interactions: content_length: -1 body: |+ event: message_start - data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01KmBXvHhii6JTJQ1oVHtd64","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":4,"cache_creation_input_tokens":155,"cache_read_input_tokens":12461,"cache_creation":{"ephemeral_5m_input_tokens":155,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } + data: {"type":"message_start","message":{"model":"claude-sonnet-4-5-20250929","id":"msg_01JPjYdy2AeAZTzdN5QA79Am","type":"message","role":"assistant","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":4,"cache_creation_input_tokens":155,"cache_read_input_tokens":12146,"cache_creation":{"ephemeral_5m_input_tokens":155,"ephemeral_1h_input_tokens":0},"output_tokens":1,"service_tier":"standard"}} } event: content_block_start - data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } + data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""} } event: content_block_delta - data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } + data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"Done"} } event: content_block_stop - data: {"type":"content_block_stop","index":0 } + data: {"type":"content_block_stop","index":0 } event: message_delta - data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":4,"cache_creation_input_tokens":155,"cache_read_input_tokens":12461,"output_tokens":4} } + data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"input_tokens":4,"cache_creation_input_tokens":155,"cache_read_input_tokens":12146,"output_tokens":4} } event: message_stop - data: {"type":"message_stop" } + data: {"type":"message_stop" } headers: Content-Type: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 5.472855042s + duration: 2.133299542s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/bash_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/bash_tool.yaml index 8cb99f7734648bf7171256a53aef339e6afce2b4..3b32647612abc769fa17989735a06289ad986d72 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/bash_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/bash_tool.yaml @@ -24,27 +24,23 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"pGHQqD7nA6xClq"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"CV71sq1ey7AbB8"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":"Create"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"TU3FQzFuok"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Create"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"8xNVXRX1xn"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" File"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"jOCPOsaRvNw"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" a"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Ck2fARrq1Zcdga"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" test"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"dskt2DKShr2"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" File"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"6BYdLE1PQhP"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":".txt"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"YQ0VBHokFitf"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"uZIMGKPzgBo"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"rU4NO4nlo8t"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Bash"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"RLXyUFVVaGw"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Hello"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"IUUse6Cjpj"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Command"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"lVssu3ZG"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Bash"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Vk3KtkrYm5U"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"r33EF1Lm9n"} - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Content"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"tgSwhxTx"} - - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"rO8vPwWjhX"} - - data: {"id":"chatcmpl-CVfMXnD4YmHCkrpioDkYdzv2CtlTy","object":"chat.completion.chunk","created":1761663201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[],"usage":{"prompt_tokens":139,"completion_tokens":8,"total_tokens":147,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"LKNzTKqrDa3ziJ"} + data: {"id":"chatcmpl-CVhDh7p7F3rfNemU8Mmbm7LH6gcYn","object":"chat.completion.chunk","created":1761670341,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":139,"completion_tokens":6,"total_tokens":145,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"aKjiwaKMTulRmA"} data: [DONE] @@ -53,15 +49,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 571.266458ms + duration: 979.4655ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44160 + content_length: 42943 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -77,103 +73,103 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_erkF6IfOlov2o4mh61BZURul","type":"function","function":{"name":"bash","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"8AI10ylDQr9A"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_083nCTGXCt06oms0XPapDQVz","type":"function","function":{"name":"bash","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"SZU7tBC4mBVE"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"lyLjYLsah3knXhL"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"a9fTGmAJPT87WEF"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"command"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7E0UBWSHDVr"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"command"}}]},"finish_reason":null}],"usage":null,"obfuscation":"EvopGpPSCUT"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"v6zLCNQL4MvKd"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"IsZjpDOJkFcew"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"printf"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CyG9VaqwdO3q"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"printf"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YHKn5RrsCoX2"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" '"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" '"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BgFO6h569qqt9"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zMgisWpv7GnPy"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" bash"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oMYv0D4W0ADFi"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" bash"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kCOV40MzERbtq"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"'"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"'"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" >"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" >"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" /"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" /"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TbzR8EYrqTNUP6X"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UZmAt11cU3xIWQ6"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GH7ddrVAERhBi5P"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2qddsOhbychwjPY"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zHIfsxnoYmjFQ9k"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cEMc1bj4ZJyUU9t"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0O517MLyFMXSJ"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Sqp6XuxGewv9h"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7gtyTGhkfrZgv"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HsvItgqI6hpv6"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GJewD53oV6g34"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"mgRbsdgZrLnX2"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gfCv9U0AKE7An"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CjvcJVXj8xNz1"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cOfbsrqjbCfAU"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kOJfElNA2T2GP"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"V"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"L"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/bash"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nD0PCY5PoVWB0"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/bash"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TD89Cst7KhbA7"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GT2kKFX9Ne7VF"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hj4huBvIGUVdz"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z7qgNBhw3fKZY"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8ANIRkLueNnKu"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"DC01jBstBSEuRx"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5EAcKhzgWcO6KZ"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"MgH6wO5V1n9Co"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"MyciMzf0tL3kp"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"description"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OTGLras"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"description"}}]},"finish_reason":null}],"usage":null,"obfuscation":"k1cf2nR"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"E1Lzhc4GMdyKz"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"C4hZ3hJYgEfdQ"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Create"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0SdJupVSMKXt"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Create"}}]},"finish_reason":null}],"usage":null,"obfuscation":"RhRYHfdoNKen"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"DW0cr4FA5o6wM"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9Wyz5OeGbQMoi"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jynRiI0IbTKIWR"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kHjWLR26jAoetP"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" with"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oSGTE4PDlGmtd"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" with"}}]},"finish_reason":null}],"usage":null,"obfuscation":"aaBzaXqj0N3Uc"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" exact"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fjoMXLtKriJk"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" exact"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6WQRNZZ8k5HH"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" content"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oKmM5lpLa8"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" content"}}]},"finish_reason":null}],"usage":null,"obfuscation":"c8QRHo25eB"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" '"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" '"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CkVI3wuKAcSUz"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jKZchSeMr3MzG"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" bash"}}]},"finish_reason":null}],"usage":null,"obfuscation":"VzNpQDZ0CsP3O"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" bash"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uKnuGeqbcpxCY"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"'"}}]},"finish_reason":null}],"usage":null,"obfuscation":"X"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"'"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" ("}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" ("}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"no"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"no"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" trailing"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6z6Lo34EB"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" trailing"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oqIpHnLKy"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" newline"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xLaDSMcyHP"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" newline"}}]},"finish_reason":null}],"usage":null,"obfuscation":"J9Z7rBatlh"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":")."}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"h"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fNdVcyiq90IVgF5"} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"e7vyE2jR2F0X7dW"} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMYBUnIg1xyd3Kt2QgwTKWEaqKF","object":"chat.completion.chunk","created":1761663202,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9744,"completion_tokens":576,"total_tokens":10320,"prompt_tokens_details":{"cached_tokens":9728,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":512,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} + data: {"id":"chatcmpl-CVhDhmY2wljzhzbtiCTfUg9gylFBT","object":"chat.completion.chunk","created":1761670341,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9493,"completion_tokens":896,"total_tokens":10389,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":832,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} data: [DONE] @@ -182,15 +178,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 11.110321292s + duration: 14.814803166s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44573 + content_length: 43355 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_erkF6IfOlov2o4mh61BZURul\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"printf 'hello bash' > /tmp/crush-test/TestCoderAgent/openai-gpt-5/bash_tool/test.txt\\\",\\\"description\\\":\\\"Create test.txt with exact content 'hello bash' (no trailing newline).\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"call_erkF6IfOlov2o4mh61BZURul\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_083nCTGXCt06oms0XPapDQVz\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"printf 'hello bash' > /tmp/crush-test/TestCoderAgent/openai-gpt-5/bash_tool/test.txt\\\",\\\"description\\\":\\\"Create test.txt with exact content 'hello bash' (no trailing newline)\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"call_083nCTGXCt06oms0XPapDQVz\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -206,13 +202,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMjv9q58WKu4LWz0oTmThQWAGGy","object":"chat.completion.chunk","created":1761663213,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"QTabRIJUa4"} + data: {"id":"chatcmpl-CVhDxPIlchOaDprEUo3YYCxK8GY94","object":"chat.completion.chunk","created":1761670357,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"pqoYMZHyOu"} - data: {"id":"chatcmpl-CVfMjv9q58WKu4LWz0oTmThQWAGGy","object":"chat.completion.chunk","created":1761663213,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"PYS3PVfl"} + data: {"id":"chatcmpl-CVhDxPIlchOaDprEUo3YYCxK8GY94","object":"chat.completion.chunk","created":1761670357,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"jc55muLr"} - data: {"id":"chatcmpl-CVfMjv9q58WKu4LWz0oTmThQWAGGy","object":"chat.completion.chunk","created":1761663213,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"adTrxF"} + data: {"id":"chatcmpl-CVhDxPIlchOaDprEUo3YYCxK8GY94","object":"chat.completion.chunk","created":1761670357,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"MpBcWG"} - data: {"id":"chatcmpl-CVfMjv9q58WKu4LWz0oTmThQWAGGy","object":"chat.completion.chunk","created":1761663213,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9814,"completion_tokens":4,"total_tokens":9818,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Oc3H7"} + data: {"id":"chatcmpl-CVhDxPIlchOaDprEUo3YYCxK8GY94","object":"chat.completion.chunk","created":1761670357,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9563,"completion_tokens":4,"total_tokens":9567,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"EQQSW"} data: [DONE] @@ -221,4 +217,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.639221417s + duration: 1.51695825s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/download_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/download_tool.yaml index f1914d3c051b057e494dcdba60c8b911d4bdea5b..f255ee21812165e9654dd887bb814f568a95d69d 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/download_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/download_tool.yaml @@ -24,25 +24,25 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ahfxGvHErD1qIt"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"7J6Ag3utyJo6br"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":"Download"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"jDkgqHhd"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":"Download"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"iINn8gZB"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" and"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"MhAVGcg6FiAo"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" and"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"lhxPetQGa6EL"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Save"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"aQ4XugXWHQ2"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Save"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"igtSKFeyoG1"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" File"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ttJfW3jOBhc"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Text"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"skCMeYxEQYD"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" from"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"WU6XYRsVeuQ"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" File"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"mt67GdKRgAn"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" URL"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"h6pKez6HGVw2"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" from"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Y5jePcwGizq"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" Example"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"eAMtRY5u"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{"content":" URL"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"7BV9row5ujHP"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"bS7B6xPAWS"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"Ursq0nK6ur"} - data: {"id":"chatcmpl-CVfMlcprQyIALVRXktkMgoAgnHvJS","object":"chat.completion.chunk","created":1761663215,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[],"usage":{"prompt_tokens":148,"completion_tokens":7,"total_tokens":155,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"0PqKx2wuUpgKHz"} + data: {"id":"chatcmpl-CVgvJQNW17UH66ulPZDk4wAJ4LWq8","object":"chat.completion.chunk","created":1761669201,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_cbf1785567","choices":[],"usage":{"prompt_tokens":148,"completion_tokens":7,"total_tokens":155,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"PeyVhG4jmpJWee"} data: [DONE] @@ -51,15 +51,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 603.351125ms + duration: 963.936667ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44213 + content_length: 42996 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -75,97 +75,97 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_HXGiXrTSOjMiG74Jp0oqHzex","type":"function","function":{"name":"download","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"pf3y1VLy"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_EPQpcr7uWuVXRDrRIg35VazP","type":"function","function":{"name":"download","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"HITW0IsU"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"wtSNo1MTFuBHHz3"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"6C7gCMWW1EtNXR9"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qUeV6qi8FsW6vf"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FSgBpZ93noA2f5"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3CgaCEYVT4jmH"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UDgfFc7q2vhWp"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5mp1SUOLhhffFTc"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jOAuQcU72EGG4LT"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yXXWbjA7T2jk2WC"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3uJ5GN4frXa0TG9"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ibDRB2oQKhRNNji"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"SS9ahpkxsNFETRL"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uKuumgc0eKr3tBL"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"P2pVTzGUpZ8V3BO"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Dsc03o1mbPl9L6t"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FiHs7MiRCgigwDF"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3OAwP0V3ujSMf"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zaSFVdp1Q3lpS"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ucAk1sr7Bv6pS"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M8XR26igiYaNm"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uTT7C79pgfEuz"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"svkAbWf8zznlL"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hi3iLAsGexuDa"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BnaqPZvBZ1PWf"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FVm0JT3iwk8o6"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nUYhKkEbBaJHb"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"H"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"U"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"y"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/download"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UNK4aq2jr"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/download"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bXohbWacI"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"og4gDdwC5cR1J"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UP5lnLc2ILH2t"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0lZOdbBYKU"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example"}}]},"finish_reason":null}],"usage":null,"obfuscation":"SiVL4yEbwg"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TNNtvYASbmnjrX"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MCYHUl3E6lWoyU"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"C31rFHmwlp0UL"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"8VHIvVWv2AMGY"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"url"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tzF0RIgh2DyV0E5"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"url"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YjJ4CiXlf8dpPfv"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"dUKYRfsD7SLRq"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"nkQENn2W524QA"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"https"}}]},"finish_reason":null}],"usage":null,"obfuscation":"x25HICOCEsj2B"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"https"}}]},"finish_reason":null}],"usage":null,"obfuscation":"t4WKZM0ag6FeT"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"://"}}]},"finish_reason":null}],"usage":null,"obfuscation":"KpPttUhCdDvY645"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"://"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OC1AiDjHTBI9wVI"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"example"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LXcmgpxveaH"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"example"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rdSziE72WAR"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FCx9vTM0j6mu"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-files"}}]},"finish_reason":null}],"usage":null,"obfuscation":"KSR5qlH6iMeU"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online"}}]},"finish_reason":null}],"usage":null,"obfuscation":"aYCxalKiiuD"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".online"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pJ8PCC1xecW"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-con"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FZuHPu4oCdmvrr"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-con"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Qqg2LhBkK4FqPA"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"vert"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GY7ActYZG5hmhu"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"vert"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hQfY8BU419ejo8"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".com"}}]},"finish_reason":null}],"usage":null,"obfuscation":"PVtNGKWUYik2aP"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".com"}}]},"finish_reason":null}],"usage":null,"obfuscation":"52OabauxKBVhYi"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/document"}}]},"finish_reason":null}],"usage":null,"obfuscation":"I18AWfJ9K"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/document"}}]},"finish_reason":null}],"usage":null,"obfuscation":"PZvVaCZL1"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/t"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/t"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"xt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"xt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Nfx8aiIDzI"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/example"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ricVSvyAIi"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"npa7Rq5ZkbZlKU"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".txt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZEVdzFgkbFj54b"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"bps5wkvSEb1EJ"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"HzMpAF6lnQ8az"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"timeout"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QH6s86TdiYJ"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"timeout"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D11o467uw6C"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zBQQJbbazCC6Wbe"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YymgcWhP7z9VxK7"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"120"}}]},"finish_reason":null}],"usage":null,"obfuscation":"N0aBgv89DFj3rSW"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"120"}}]},"finish_reason":null}],"usage":null,"obfuscation":"eSCKnoU8mCJ80gD"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"j"} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"o"} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMl5qA9JKKhCkF0d5TgKzGzT2oR","object":"chat.completion.chunk","created":1761663215,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9752,"completion_tokens":316,"total_tokens":10068,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":256,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} + data: {"id":"chatcmpl-CVgvJlRgHSesLLh1jVhN5KqfCXXvV","object":"chat.completion.chunk","created":1761669201,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9501,"completion_tokens":508,"total_tokens":10009,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":448,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} data: [DONE] @@ -174,7 +174,7 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 7.383364042s + duration: 7.725510667s - id: 2 request: proto: HTTP/1.1 @@ -220,15 +220,15 @@ interactions: - text/plain; charset=UTF-8 status: 200 OK code: 200 - duration: 136.773459ms + duration: 214.7535ms - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44760 + content_length: 43543 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_HXGiXrTSOjMiG74Jp0oqHzex\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool/example.txt\\\",\\\"url\\\":\\\"https://example-files.online-convert.com/document/txt/example.txt\\\",\\\"timeout\\\":120}\",\"name\":\"download\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"call_HXGiXrTSOjMiG74Jp0oqHzex\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_EPQpcr7uWuVXRDrRIg35VazP\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool/example.txt\\\",\\\"url\\\":\\\"https://example-files.online-convert.com/document/txt/example.txt\\\",\\\"timeout\\\":120}\",\"name\":\"download\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/openai-gpt-5/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"call_EPQpcr7uWuVXRDrRIg35VazP\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -244,19 +244,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"yL8zjfSPVO"} + data: {"id":"chatcmpl-CVgvRcyCtPdmxcvLwv7LagL8HHmCW","object":"chat.completion.chunk","created":1761669209,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"DyyKfTOVeL"} - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Saved"},"finish_reason":null}],"usage":null,"obfuscation":"Hdz16XI"} + data: {"id":"chatcmpl-CVgvRcyCtPdmxcvLwv7LagL8HHmCW","object":"chat.completion.chunk","created":1761669209,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"WMtu5Gcq"} - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" to"},"finish_reason":null}],"usage":null,"obfuscation":"ci3kMjZpB"} + data: {"id":"chatcmpl-CVgvRcyCtPdmxcvLwv7LagL8HHmCW","object":"chat.completion.chunk","created":1761669209,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"njy1Q8"} - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}],"usage":null,"obfuscation":"DyRk"} - - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".txt"},"finish_reason":null}],"usage":null,"obfuscation":"mVmXzeW8"} - - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"VWGH2J"} - - data: {"id":"chatcmpl-CVfMtftzW0MAM08U2NaPw9vsH9dy8","object":"chat.completion.chunk","created":1761663223,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9853,"completion_tokens":7,"total_tokens":9860,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"NeSTD"} + data: {"id":"chatcmpl-CVgvRcyCtPdmxcvLwv7LagL8HHmCW","object":"chat.completion.chunk","created":1761669209,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9602,"completion_tokens":4,"total_tokens":9606,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Xjs3i"} data: [DONE] @@ -265,4 +259,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.38613725s + duration: 1.828980667s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/glob_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/glob_tool.yaml index ab1b96f1b30e135908ad5b8463e2dc1a3c2492e6..69f98ac019558b33b15e9360f2763723a7a6cd7c 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/glob_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/glob_tool.yaml @@ -24,29 +24,29 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"VdRj2QwhmU7ylH"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"XT9GqzDU0960Vx"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Finding"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"C4EScgKIl"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Finding"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"uslINOnZH"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" ."},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"1g1no7aoF1lit0"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" ."},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"vVswcl24zmWoFd"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Nw1FvrTnPnF2ll"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"n7tN3f8v2hoS9d"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"1krebXYUh1"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"sZ9BxgWPh1"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"SFb7m1CZGJTq9"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"KyPLrdTeGq1"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Current"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"dAiDQ8tR"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Glob"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"bKD5zdL2MWE"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Directory"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"07u1V1"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"wzcXnQY5uvrFS"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"WYSkQQpiPVu"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Current"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"mZnx1WyL"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Glob"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"nNMNWG4alcm"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Directory"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"h8vvmN"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"BgkEiRn2hp"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"diReY6tThA"} - data: {"id":"chatcmpl-CVfN4pzznFb333KUwRukm5TGHmkJ0","object":"chat.completion.chunk","created":1761663234,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":137,"completion_tokens":9,"total_tokens":146,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Mrdw9wkK9yzDx0"} + data: {"id":"chatcmpl-CVgvUNPHYkWlpkwJK8L7aItwE4kIh","object":"chat.completion.chunk","created":1761669212,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":137,"completion_tokens":9,"total_tokens":146,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"xqfugmoKOj4QCo"} data: [DONE] @@ -55,15 +55,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 602.034667ms + duration: 1.707555042s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44149 + content_length: 42932 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -79,23 +79,63 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_seU3vSJ9Yo8NZXxqRfQulHrr","type":"function","function":{"name":"glob","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"drKrcoH3xMiJ"} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_81ojWAiDKj6brbxF9i5naNN4","type":"function","function":{"name":"glob","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"LOAKWHU2GmAp"} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"OssHoGVhB2TLywv"} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"xejEhOtAgYKoQqz"} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pattern"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A6GHaBDw70F"} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pattern"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FOmSxsty2oq"} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"gIfVwYgAE1ptF"} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"XOs7SPmqa4fpj"} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"*."}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"*."}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CfLeBQuYNbO9VxN"} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"fTLjHDm4tZG81"} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BDSdfXZ6UgQwYm"} - data: {"id":"chatcmpl-CVfN5JCQ3mJbIuT5TtwRWxkxdHFYL","object":"chat.completion.chunk","created":1761663235,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9742,"completion_tokens":87,"total_tokens":9829,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":64,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"57smxl"} + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dQNlAb9U6y8PVIq"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"H7X6Io4SKCiZJOE"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"y8NHFDDhrb5FEPX"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6j9kuoG19aPGIeu"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qMv1vTMZmjvbA0r"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z1OPcAeeq7oTL"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dzAZTNXf1VZ4G"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"t3LAZj3aFRx27"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0RhRXTIGA4zyP"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2VOjRboTgToHw"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"G"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"K"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/gl"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wmZGJKwZtBZgRIu"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ob"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HqHLVhGNWnVdy"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"h8xUWDh5XPDJnFT"} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgvUOAjiAAszXOZPxSP8kDXr00XS","object":"chat.completion.chunk","created":1761669212,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9491,"completion_tokens":107,"total_tokens":9598,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":64,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"8R"} data: [DONE] @@ -104,15 +144,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.828790041s + duration: 4.823159625s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44443 + content_length: 43293 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_seU3vSJ9Yo8NZXxqRfQulHrr\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool/main.go\",\"tool_call_id\":\"call_seU3vSJ9Yo8NZXxqRfQulHrr\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_81ojWAiDKj6brbxF9i5naNN4\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"*.go\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/glob_tool/main.go\",\"tool_call_id\":\"call_81ojWAiDKj6brbxF9i5naNN4\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -128,49 +168,49 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"9O3A8fJ9O1"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"XlohVCWvNm"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"8h7MEBbUkqF"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"Y9NjSGl2aqS"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"9TAvMlT1m"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"6z2DApfQw"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"HchdihQ87"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"MYNUf1RFR"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"XhEIvLIFX"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"H1fgGF0cX"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"ecXr2Zv"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"PwOuHqa"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"BRsuYYS"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"KXNH0G9"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"jYApV9X"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"DziP8x9"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"FVK87Vp"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"M3xCkMK"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"q35DSKS"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"ZzOzz4F"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"R6v9kM9Sey"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"frVZJroG1n"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"zcXUCaflFS"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"csaU6SR1vZ"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"v3O7TS6ITN"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"xdgIlQJLo5"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"HutD9ycN83B"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"rAF6eB9vdxY"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"M5DpwBWwonx"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"4wdMCXaqyr9"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/gl"},"finish_reason":null}],"usage":null,"obfuscation":"BFnTTLMQb"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/gl"},"finish_reason":null}],"usage":null,"obfuscation":"PSvleZjtT"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ob"},"finish_reason":null}],"usage":null,"obfuscation":"4TIyMhb1pI"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ob"},"finish_reason":null}],"usage":null,"obfuscation":"bkzkAGA6rC"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"e8FTZmU"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"RvNd8m6"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"cOj2vOV"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"P2G4Nie"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"2aGZFSY12"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"Ujykoi5C0"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"bfGIgE"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"w7gKcH"} - data: {"id":"chatcmpl-CVfN8ZnNwFvSVbado9rW4SPqEMbsV","object":"chat.completion.chunk","created":1761663238,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9787,"completion_tokens":22,"total_tokens":9809,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"nKE5"} + data: {"id":"chatcmpl-CVgva9vDxJJ7g61nx114fHIvMsbcX","object":"chat.completion.chunk","created":1761669218,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9556,"completion_tokens":22,"total_tokens":9578,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"NkbX"} data: [DONE] @@ -179,4 +219,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.663874667s + duration: 3.636113459s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/grep_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/grep_tool.yaml index 57ad846aaa320001fbf799f86981ed7b900be35e..6839ff99e3b2e3e68948e6b2d0dd0454ea08b13d 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/grep_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/grep_tool.yaml @@ -24,31 +24,33 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"uXg68pOTl3PHAu"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"hwWQ83PzlU0YFS"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":"Searching"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"nvKUZDn"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Using"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"1xnoz1Hqvfr"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" for"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"LR1rPzLCwejn"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Gre"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"6Nl67AagK08G"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" '"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ydrFFJHYPHERIG"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"p"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"xfl09QHIKNUGhdw"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":"package"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"s8VQ3GK11"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" to"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"cyR5rpi8CmF3V"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":"'"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"0v5a1A9LGYQ72oi"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Find"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"9g9CgO0mcfz"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"z1k9cJQ5P8NAx"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" '"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"M9TulzZYjjhuK9"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" Go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ek1lU1KOtB4IU"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Package"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"1dGN4FpgH"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"jZt2RbFyNS"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"'"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"kLwVXUMEezscjBP"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" using"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"X2tWIoxtnF"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"zCHyWSB72kiJH"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" grep"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"BNYVfD20VtX"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"4ymBU1Aql1CRF"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"t4C3bCA3xf"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"XOxQjy2GSi"} - data: {"id":"chatcmpl-CVfNBlhP8asPAxxVynV8y1PPY3nsu","object":"chat.completion.chunk","created":1761663241,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[],"usage":{"prompt_tokens":138,"completion_tokens":10,"total_tokens":148,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"0xwaEcHF5c6ez"} + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"7lIoElixLB"} + + data: {"id":"chatcmpl-CVgvf70TpX4BoJLvvicye473NVp0D","object":"chat.completion.chunk","created":1761669223,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":138,"completion_tokens":11,"total_tokens":149,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"t4UNaDOJdykiT"} data: [DONE] @@ -57,15 +59,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 644.30575ms + duration: 2.541330792s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44147 + content_length: 42930 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -81,41 +83,41 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_wgKdeovcswRIXePACmqckpL3","type":"function","function":{"name":"grep","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"Fwz2WK7dDw6W"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_NoF69cgWEqNMQAdDz6edsAN0","type":"function","function":{"name":"grep","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"9yjxvg73S9yv"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"tM0SoPeUFHUIjRY"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Y7hW2bKn6JHXYPY"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pattern"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tUteBt2lsMp"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pattern"}}]},"finish_reason":null}],"usage":null,"obfuscation":"eBU4sx93QLs"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"K87SL16j5kcfN"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"fhUCcQFeB8rA3"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"package"}}]},"finish_reason":null}],"usage":null,"obfuscation":"prpzfYQJUdM"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"package"}}]},"finish_reason":null}],"usage":null,"obfuscation":"PoYux4If62h"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"aOaQa0PaCKmsi"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"KeoJjAnPPSSqY"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"include"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QvqZBgSGwB9"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"include"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qALwN959MPW"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"AFdQtUUpJRALT"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"XHoYMybVVdLka"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"*."}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"*."}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"IsYwnccgGvYVh"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"n1YpfnCuCQjjz"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"literal"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hwLubgKPQit"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"literal"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NC4j7ySysJ3"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_text"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gDZSg7Pvjy73B"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_text"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Smcufw0tTyAka"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D0hX8DAPA3sZNBx"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OZ7MgW3cGoHSJSc"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"true"}}]},"finish_reason":null}],"usage":null,"obfuscation":"voSAO26GUSxMpU"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"true"}}]},"finish_reason":null}],"usage":null,"obfuscation":"whCtCag3mgxFHP"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"P"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"J"} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNBc0TDWVkw8Go9MSyubBcQxXfj","object":"chat.completion.chunk","created":1761663241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9743,"completion_tokens":225,"total_tokens":9968,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":192,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"H"} + data: {"id":"chatcmpl-CVgvejWIXsoTfgkHcXOvu7JGGCHdr","object":"chat.completion.chunk","created":1761669222,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9492,"completion_tokens":289,"total_tokens":9781,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":256,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"u"} data: [DONE] @@ -124,15 +126,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.6056395s + duration: 6.1928795s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44539 + content_length: 43322 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_wgKdeovcswRIXePACmqckpL3\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"package\\\",\\\"include\\\":\\\"*.go\\\",\\\"literal_text\\\":true}\",\"name\":\"grep\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/openai-gpt-5/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"call_wgKdeovcswRIXePACmqckpL3\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_NoF69cgWEqNMQAdDz6edsAN0\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"package\\\",\\\"include\\\":\\\"*.go\\\",\\\"literal_text\\\":true}\",\"name\":\"grep\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/openai-gpt-5/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"call_NoF69cgWEqNMQAdDz6edsAN0\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -148,63 +150,61 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"HOGYv1OVnU"} - - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"FCb8uPehGGm"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"YU6S6Bjr4l"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"uve4jDrxL"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"SPOzX5R8dTw"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"mFvJgSRFh"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"OPEPEmzRI"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"zOIvf9jS3"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"IoO0K4Ee0"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"wFQ7Ryg"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"LmDV9YdoC"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"0KRUBRG"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"2Dka7Yc"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"HotTBIk"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"NgBC0qp"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"4oNviqb"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"k27gq56"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"BlXYurf"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"ya7Zb04"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"jCUgapiZ4f"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"GNRGQcx"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"YePsk60wTH"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"B8ogA56Jaz"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"2ns7OkymlN"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"oxRdJD1ScZ"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"5rumsvWDy1d"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"S2MLl2BGkX"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"rGWmHJ3xpee"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"kRzmXYBcqn6"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"qQP0roMhXqa"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"Bsm9JCnEtHL"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"grep"},"finish_reason":null}],"usage":null,"obfuscation":"AEUSrE2n"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"AK3hjCrNEG0"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"h3vRUpH"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"grep"},"finish_reason":null}],"usage":null,"obfuscation":"Hjq7HVUb"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"sPT4DeG"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"epbK7uc"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"aRmU0Ai7v"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"lLMZ2Wg"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"APv5jwHmaF3"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"OxInYDoxq"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}],"usage":null,"obfuscation":"llXJIMJTcqF"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"Js5IMLTd8ZY"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"xl8ZBUmdhYv"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}],"usage":null,"obfuscation":"MW573m1T6r8"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}],"usage":null,"obfuscation":"z0pxvrQWXbr"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"bxTnGIX1Kc2"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"T97KtlywbVG"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}],"usage":null,"obfuscation":"01eo8ecG1o3"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" package"},"finish_reason":null}],"usage":null,"obfuscation":"deN5"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" package"},"finish_reason":null}],"usage":null,"obfuscation":"zRZv"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}],"usage":null,"obfuscation":"YjlMy9W"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}],"usage":null,"obfuscation":"fsso101"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"a0Blok"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"wi28MA"} - data: {"id":"chatcmpl-CVfNFvfHLLsLYFXtbsOHKiwN7DYs2","object":"chat.completion.chunk","created":1761663245,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9816,"completion_tokens":29,"total_tokens":9845,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"2e2R"} + data: {"id":"chatcmpl-CVgvl8QQaQGHdKvnY6f3rMEyf3QmD","object":"chat.completion.chunk","created":1761669229,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9565,"completion_tokens":28,"total_tokens":9593,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"znM0"} data: [DONE] @@ -213,4 +213,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.415104709s + duration: 2.403646208s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/ls_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/ls_tool.yaml index eed52ec950ed064fda2ac19ccd4ca48a26b637b2..c618a630f88773e0ef9d94e6ee6133310067ad84 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/ls_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/ls_tool.yaml @@ -24,27 +24,25 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"yeVbLsn0KSfT9A"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"gNkL62dKgy73nV"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Listing"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"HdY0E3Ldb"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Listing"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"KCYCOvtXB"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"w69gt8VUsl"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"uY0wbfbuuU"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"mLypgnOvOoH"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"L0aYYnHh4GH"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" the"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"xFUxzyiq5Pe8"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" '"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"RTS3q0hfHxzzXE"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" '"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"VszhbgLYgsCso1"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"ls"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"waFr55MVIusmkD"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"ls"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"SMM88sffvdcVCV"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"'"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"QKf4wqyaneRAHft"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"'"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"2u9Uj0ZgM9LEE2z"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Command"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"KcdzXh6w"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Command"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"IWzDxfqt"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"EuevlKWvGM"} - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"wScu5JD1yH"} - - data: {"id":"chatcmpl-CVfNIb8BlKZf39oxwxfCKy481Iozr","object":"chat.completion.chunk","created":1761663248,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":135,"completion_tokens":8,"total_tokens":143,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Scr3c5WYWwoc8v"} + data: {"id":"chatcmpl-CVgvprMBgmBk4sF9BSY3FlCRjRVrj","object":"chat.completion.chunk","created":1761669233,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":135,"completion_tokens":7,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"tg68EKKLKWloVt"} data: [DONE] @@ -53,15 +51,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 577.433417ms + duration: 1.675369791s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44141 + content_length: 42924 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -77,21 +75,31 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_tsUghih6YJiT7VDMjcKKymFu","type":"function","function":{"name":"ls","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"6bfZOIowgje9BC"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_0slM0J7jFWeEWXQfN56PRsy7","type":"function","function":{"name":"ls","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"cFWJGksjDujhaO"} + + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"u4l1hfRZFm6ebQ7"} + + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T0eBff2E42x7H1"} + + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jz4RaJ4hnR4TtRL"} + + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \".\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"z28fyxTIwme"} + + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"NWQ2Cco9HAoAMMD"} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"GoItwxJsnUXqO4H"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"depth"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9M8nJz3pd00e9"} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OCYtCse7uz45Qh"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FtcG7s87HpxK4kz"} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wMw0zcdvnGXkO8s"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"R"} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \".\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"hXEQ6F4YyxrR"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D"} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"C"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"e"} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNIW4BWNW8k0ptxWXv4mVlAgDq0","object":"chat.completion.chunk","created":1761663248,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9740,"completion_tokens":215,"total_tokens":9955,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":192,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"i"} + data: {"id":"chatcmpl-CVgvoARKRTZdbjjz0EjpnDeIEOlTQ","object":"chat.completion.chunk","created":1761669232,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9489,"completion_tokens":284,"total_tokens":9773,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":256,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"7"} data: [DONE] @@ -100,15 +108,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 4.924636s + duration: 7.627417375s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44450 + content_length: 43247 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_tsUghih6YJiT7VDMjcKKymFu\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\".\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_tsUghih6YJiT7VDMjcKKymFu\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_0slM0J7jFWeEWXQfN56PRsy7\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\".\\\", \\\"depth\\\": 1}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_0slM0J7jFWeEWXQfN56PRsy7\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -124,21 +132,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"WTdxs8UUMN"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"PiVcTTvjny"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}],"usage":null,"obfuscation":"a6on38LjPj"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}],"usage":null,"obfuscation":"cOv1F01PqH"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}],"usage":null,"obfuscation":"CJ8t7UFQ"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}],"usage":null,"obfuscation":"b6h3T4Tk"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":","},"finish_reason":null}],"usage":null,"obfuscation":"PODmRDaYJJn"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":","},"finish_reason":null}],"usage":null,"obfuscation":"B9sRYCdMnPI"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}],"usage":null,"obfuscation":"B5K4ItU"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}],"usage":null,"obfuscation":"Q24WoZI"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"AX2ULhcFw"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"31LkttrwY"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"aqVOUD"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"NjaQpT"} - data: {"id":"chatcmpl-CVfNNT2rvCBxu9dLkmLQK2LPPRtkc","object":"chat.completion.chunk","created":1761663253,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9797,"completion_tokens":8,"total_tokens":9805,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Rjc1W"} + data: {"id":"chatcmpl-CVgvxzjeMFsbswL0T15wtgCM7F83W","object":"chat.completion.chunk","created":1761669241,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9550,"completion_tokens":8,"total_tokens":9558,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"kHMz2"} data: [DONE] @@ -147,4 +155,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.170903042s + duration: 1.627885666s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/multiedit_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/multiedit_tool.yaml index 612d39d3224e7dbee4ab349505ccbabd2b3ef702..f52ce79bfbc1db336b8483e592692b326806d7e0 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/multiedit_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/multiedit_tool.yaml @@ -24,31 +24,29 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ZGGEAO7nAiRHuY"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"UZFzfcRVyawjbw"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Edit"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"sGVUVk17GWiV"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Editing"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"aNYNQczlb"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" and"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"4EjiKJ0CABCD"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Code"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"xm6vKiBaN1D"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Comment"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"59fun9hp"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Ftb6lWyFAbf"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"lbifuILM5ZlOF"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Mult"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"NvFXDA63orX"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" main"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Eantz1xKbD7"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"ied"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ztlgWDGgj0ext"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":".go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"sg13lSyn4a1HI"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"it"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"0kzJh0U0o16rJ9"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Using"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"cTsgnTqcac"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"nvVvkYM32B3lF"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Mult"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"bEHC2WPeTO9"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" main"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"gXqFH6e3T2K"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"ied"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"e0gNGHeQXHLxM"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":".go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"I8RcCsuuOZdbh"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"it"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"5U5ntmSTK5VYeB"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"lvinTEjQWu"} - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"6WmumR0z8U"} - - data: {"id":"chatcmpl-CVfNPVIBrpBNuB2oGNCxrKbRgYnGx","object":"chat.completion.chunk","created":1761663255,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":157,"completion_tokens":10,"total_tokens":167,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"mu62hWyLhZ6Da"} + data: {"id":"chatcmpl-CVgvzBWdIz2Ap1hQF4lRmc6MnGLC8","object":"chat.completion.chunk","created":1761669243,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":157,"completion_tokens":9,"total_tokens":166,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"hSkJt0qOmvTSML"} data: [DONE] @@ -57,15 +55,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 579.264458ms + duration: 928.928667ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44227 + content_length: 43010 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -81,71 +79,69 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_JlXpqBYdqtlfFG5JZWSEbRoo","type":"function","function":{"name":"glob","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"jeDpgg7JqGS2"} - - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"TcarHRjfkvaWplL"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_1EZWwA7F6vCnLSswkrzCH0G9","type":"function","function":{"name":"ls","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"nZh2suTBH5oezq"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pattern"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IjGeesFYp2o"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"RsFVoaSfrL43Tjq"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"VqQPszhCmMnwZ"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9TKL3FF7oA1hmi"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"**"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IOW3drjUiKTVgcs"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"4"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LGEkjS0XemOwjx"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ASOlLmbsY5VnYK"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qiZe9XAMeTolJ9Z"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"mSVPQxJv9jPP2RZ"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"VZzUMn7EuEXYNty"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"ej1da6Fxh7WOk"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CJr97ggMjAj3iCs"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"RRrHKkNfGXKBYW"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AMVYCYOBVOdjA"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"a7tD8w6LAkv6T9K"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cINhAJhCa6izl"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WwaT90mMiFFlSHi"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"RKA1QfGklVAJs"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kdgdacukmOhti9w"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fBX1QSAxyYYZX"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A81oMVS4x1kZwbk"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AZKsTrbMqecEc"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YFrALp0x2OkAUAc"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"N2PCUSmuRkxLX"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZxV6XGBIr3SmS"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pTBAkA5hEG3GU"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Y"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nbri29P8fjIhx"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"h"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1GA8IinOzah1P"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2N9JAfcQkpAycyP"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GH0aYwTRjN7S2uA"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"o"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qWtSxfGqYyqpJ"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"e2u3Plcy7Df01Gi"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Mcn8mQHzj4nwzB9"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YQ0UJhsJ9zGSOBO"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"depth"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5rDDPrHWI3A7y"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"I8ESlLUtoaRixWT"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"mIgLR6MC2eG7648"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"s"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Rpf9bMjlKs81q"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"2"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iDNMB0dCMtyF7kQ"} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3"} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNPqT4CapZVYABcDi4KD5r9fPGq","object":"chat.completion.chunk","created":1761663255,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9764,"completion_tokens":239,"total_tokens":10003,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":192,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} + data: {"id":"chatcmpl-CVgvz6zLSV3BmFQPba3oGL2q6q25y","object":"chat.completion.chunk","created":1761669243,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9513,"completion_tokens":175,"total_tokens":9688,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":128,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"1"} data: [DONE] @@ -154,15 +150,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 5.525106542s + duration: 4.187234917s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44604 + content_length: 43397 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\",\"tool_call_id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -178,63 +174,73 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_p6iPAE1CcqORgJwTF1Mh2Dgd","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"jjmu6PrFEiyQ"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_YF7ugjL9nlAcYgcMrSmtonJ4","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"nDirRqQn2LYm"} + + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"5DWh3IFt2Y2xXI0"} + + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"t2UOr14Zgof1ne"} + + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0QYetjPVUDiII"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"sIuWWD7xnN4KxEC"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wN6pTYPoNaeSVvb"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fAWAcit4x2bhSK"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tpdm7FIE25xBcgg"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"EqYkwd3s55dYw"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oDGqlltFug60mNl"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tl1DhkVrFvzC9o4"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FJ1FDH9dGozXdQN"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"X76YJ2pOPIUfUKJ"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NZyjVX1LFlVpLBr"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"grePMfsmAiOi771"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XEVUWEB92wD9v"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QQ0gQgZqyQGbO3o"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"KxsRv2Q4MqTSd"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"km2hvGj9Q9QjrFf"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HRSF0xzJPwYv1"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TBKhn6qnBWWFe"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6goOn2zxUtKSz"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"j7asDWoJ1OCmz"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1rfzSXgzfaSQU"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oVOXIxyTwMg9G"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pcT5Xbx8e2ujA"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CofgSEjYFf1kj"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"4"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"x"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"L"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gnw4UMlRFYMw7pn"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"f"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0KA3ZR1vuOvO98y"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5DrVvcdEmjfoRLr"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oHA1NAJKWFUpr"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wZWEuZsVmYH44WA"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6z5xYFVb49eem"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5torZcB51YphLO8"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"s8Zi861tTLffM"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"uMARhochPe3OM"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bWlYT1HBQ9yvO"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"limit"}}]},"finish_reason":null}],"usage":null,"obfuscation":"JS0W8QzDgkARp"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FBtypgddGjjAgHT"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6qEgE6pnS8UvVNO"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"79f3zVUFc8xJXRK"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"200"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Jlvpd8gISxhdLS4"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"0"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2"} - data: {"id":"chatcmpl-CVfNVhpSrhVD8WOz1zWcQVMhbWxdZ","object":"chat.completion.chunk","created":1761663261,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9835,"completion_tokens":37,"total_tokens":9872,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"3omf"} + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"z"} + + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgw7BLuyJmnQyxMY4Uv2Ph8G1Cjd","object":"chat.completion.chunk","created":1761669251,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9594,"completion_tokens":42,"total_tokens":9636,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"hOd2"} data: [DONE] @@ -243,15 +249,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.654741541s + duration: 2.321744916s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45053 + content_length: 43861 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\",\"tool_call_id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_p6iPAE1CcqORgJwTF1Mh2Dgd\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_p6iPAE1CcqORgJwTF1Mh2Dgd\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_YF7ugjL9nlAcYgcMrSmtonJ4\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"limit\\\":2000}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_YF7ugjL9nlAcYgcMrSmtonJ4\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -267,249 +273,221 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_gk3SQodQcQC8MDUFYDdkM7yt","type":"function","function":{"name":"multiedit","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"nEwWlNQ"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"0cbDlZ2mWzhyVEs"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8E0Ks0oveoqx3u"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oOUkt6Hw8nE65"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yfBTGfWr5olw2xZ"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QEsFw4WYvUGwQ7B"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0oCMAZ2r5LSNnXw"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zWupvZlSMZSh5np"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wcnyCSrw7BIUXu3"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"eU4Ae5fz1AYSK"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"s5BchkOk8Rz5Z"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_7awYAPC9lIeDqVpsfkNcGmiq","type":"function","function":{"name":"multiedit","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"3lvzmD8"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"W7alXmjA9LEcg"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"WfrI70HLwEA1dMP"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pxsHHEAeJGZT6"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"24oX8BUtRkc5mb"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oFDsWcd72QaUy"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tzxYahLdZBkVj"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lFr2WdUNtRBkHhD"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r8EbnVlHm8mAoEs"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5Q9ZsxjOeJ8vkMX"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M0H3i8gF0M04mJL"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YhnNiyPv6PELcHL"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"u1mCcsfqKqwmP"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CMViUUKnIdYxLm6"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A4CbgQZBRnleL"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jVT1kVtdUXGlwSi"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5AkBSfiF3xcS3"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MzcMUvIM64rvC"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HMtQyP7STKESx"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jG20WdSJD1CZk"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YcqJKTPyNiq8y"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FPTRXxPBE5zsiuo"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Yb18vqSRxCm3X"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ed"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"S"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"its"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nHumwWA4ROcp798"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"V"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":["}}]},"finish_reason":null}],"usage":null,"obfuscation":"nOppAspyNSO82m"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"ksmc9adshdMNfCu"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uL3QNnmXjEYhq4C"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5u5C0mRDSwvrxrT"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MJWTMvgmdnCpFZz"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ySqQg5KsWTd"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"gCWZl4whFcbLj"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8o2IqFkxeajhJ"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"4PAc9PmdrL9Lux"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ddicaGzFkC6pt"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BTz4dxsV7Vhbi"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zllmYiLhrJkX1nc"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"1xz0ioEcrgVW1"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XocRUmhqthBiVL"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ed"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"I"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"its"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bSiqbUVBCzUwrsx"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"o1DiUHciG2T1eQC"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":["}}]},"finish_reason":null}],"usage":null,"obfuscation":"TrNFJIr1shbasE"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"eKsgwL2fiVzasWM"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"y46jwGdhqBOGg0s"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"krA1yyHNC4"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9RL4xCQKkAAExNe"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"WYRa8D3h01MU3"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hbGkDNAj75I"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"X3P7S7baA3UO7"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"LtkGSQy4b516Q"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"R"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" World"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Bb7UofoLIOXu"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tf"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MVsiOAM0f67cdq9"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"mt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZajvVid7X9YmDYH"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WcUqGjBOHw"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"V8vHtSyJYWYeW"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2noiUowxLG9tF"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"j"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"f"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"fyitZ3w42ankY"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" World"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ybXHpwa6IThU"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HHVTBbLt5OoVxvo"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ldL3QZ7GYhMCY63"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"p0L1c0IarUy"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dZyC00DhlP1MkBc"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"PAp6lx2rek6zB"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"6ob1hAL7GILnN"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"du4CdqeiqtrDvI"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nb5N0eUew0wgxtW"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q9DuLY8uZNRNx"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7eaiWaFvWpX"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"OQLVn1fw8j7BX"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jp9yJMsVZ2wDPD"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"u"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tf"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YaoU2Q6jof9aLjH"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"mt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2XswdGuCLc1C5fJ"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fqjl4aaD96"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"biaKkLVqsl"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"WYIMOD2APpY8o"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"4xq9vmZ7bwN51"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yQTfMS1i1YQ4t"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9oz8jToT0RM1N"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"Y"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"j"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6e6y3BPSknXr"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GUcrt13ezosZ"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MEa9XLiIvktu3c1"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IuyH6t1GFsjXRul"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M4pTzMqe8yqO7cr"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gIEbPPmD9ZECFWG"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"},{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"QFZH52XW0EL"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old"}}]},"finish_reason":null}],"usage":null,"obfuscation":"noemMdHUFSTa4DU"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"K"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"PSgMQDmIn0A"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"iPksClRJqxBXPzZ"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"goCAClNkoxaqT"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"},{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"eY7UWUhtjikXS"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WrLpcoPbDVvadN"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qIX6YgJQtw7Ig9L"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vUeUQku3BmEdt"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pNxsYoyso9C"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"lSzDeIeAC7oo6"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QUVcJg714QPXsH"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"B0knH923ENgZPH"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hpTi3LwE2bak8"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6NtJynbF8wRPkSl"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XszcLBHLuBgSSEd"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xviaCpyxsm2AhU"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZppHVBlo8I"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"j4WYbVWDA8jLk"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"JwXoYcYB7A3LFeR"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6CZ3ajo2XV0RL"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iUmoCUUAdG3zrAD"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"F"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"E9h9lsjr2E"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ymEJesHZ7tdJ"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"668J3EjLctBny"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gAQ7rNJc0eiK40v"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GWad1gEDsQu1N"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5S7OrYfjOYiIQoz"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"t"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"l5TPF0aGMNvh"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"n3ALEOBbEstZsBb"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"K"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"i4458JWvxpwM7U2"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"xXl3IrOr9O1PO"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5nz4BfI57QpYLge"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"JV1vohRvIVY"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"L"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Ebyx0F1KXmesn"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"uGVxcZ8yeh9oS"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9xunTXuUj45ofA"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ewOtulq0ie14PKH"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YwmyhZX8nhZLU"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"trOL3fSGcPt"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"PdaVDZXJpEAUv"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"B554gwaVJuwXS8"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D6oxJqh3R78b50"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UUwtcBQsJeaKg"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2kz92l2xddQfREN"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"//"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"RoxrLNFJGTLpWf"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Greeting"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vFWDKxYHq"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"x82nXY7DPcGFG8Y"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Aslx97CUjqZFK39"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d6NvpCT7vI04q11"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"//"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"C5p5Qq6vHWZtST2"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Greeting"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WP5GERDPM"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oS8ofp4I40"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"4ZWcGx7ZJbmQUfr"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"5AOonvZZpSeaY"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"READJK2WzE8XNZH"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"t2x59Xyh8lZtY"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"aHIz4r9vmQ3c45t"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"k"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HhEgP6PMNd"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sGAxAtgxZecp"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"7uO9fycDO5jUD"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uu3g1Vd9Ohb3OUx"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q8GNVvwX0kPtr"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"z9iF5R9Rqve5uWK"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"f"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d5lrz1uT94BJ"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nvXeVFGYGw1CcyO"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Dt9M5YiRN6Sa5wZ"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"KIVMHiSr1ud8Ubi"} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"]}"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"H"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YzADHJiJzja3AAs"} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"]}"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - - data: {"id":"chatcmpl-CVfNXiasfeuZbVoYiQ2lzBllbF4qM","object":"chat.completion.chunk","created":1761663263,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9934,"completion_tokens":1418,"total_tokens":11352,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":1280,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Xce1Ce7OwKQfUF"} + data: {"id":"chatcmpl-CVgwA9IbAyA8mrSKhJ0mhwAt5cYKk","object":"chat.completion.chunk","created":1761669254,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9698,"completion_tokens":1852,"total_tokens":11550,"prompt_tokens_details":{"cached_tokens":9600,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":1728,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"AVMUHbXTRHyq3C"} data: [DONE] @@ -518,15 +496,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 19.061113s + duration: 33.294924s - id: 4 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45810 + content_length: 44578 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\",\"tool_call_id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_p6iPAE1CcqORgJwTF1Mh2Dgd\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_p6iPAE1CcqORgJwTF1Mh2Dgd\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_gk3SQodQcQC8MDUFYDdkM7yt\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"edits\\\":[{\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"},{\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"}]}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 2 edits to file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"call_gk3SQodQcQC8MDUFYDdkM7yt\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_YF7ugjL9nlAcYgcMrSmtonJ4\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"limit\\\":2000}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_YF7ugjL9nlAcYgcMrSmtonJ4\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_7awYAPC9lIeDqVpsfkNcGmiq\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"edits\\\":[{\\\"old_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\\"new_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\"},{\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"}]}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 2 edits to file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"call_7awYAPC9lIeDqVpsfkNcGmiq\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -542,63 +520,89 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_l8oMEvpMDIkvfbAq584WE9ir","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"rW9MkXvx5IBv"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_1bRloNLG6Jgx4JsfmNpQYRdd","type":"function","function":{"name":"bash","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"WnI0uRF44TMW"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"uCfAgDS9WlkSqTU"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"command"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Xlt9zF4A88V"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"MxzGH5M6aS6GH"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" run"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Y07bIVMaJck8VP"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" /"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9Y8NfoKnx86EnxO"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rlCuuuPTpT3MZBc"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rMHpfrHH93Hu9it"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yP5vyEyfeAYdX"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nRx4taJ2K80vI"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5YMunSvDrp69J"} + + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rquuA1CnCAMUU"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Fx32KCOEpfbrfBq"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"crccA0Xfbwvcp"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jwNf5YVfLW0Kae"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"L6LKuS9tjHaeZ"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GmP6Pa5hE9EV221"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8RFM3CyUSPJbgrY"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"C"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"86w2pHHiU99mfl8"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"h"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fZNCAJGcM0wsaCx"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LVxYghdhlUr7xn3"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zgEMTaiw0NzyCWs"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bMErdGJuKUHhi"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"a874AC2nDnYPOvh"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gBY12OWT0wGhP"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"JsmNfEMFZzDNg"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OpE5cnL9sIQna"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"i98bUt7VO0zcD"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fgCsmcJCwoAA3"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2rI4IznK84NSF"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tWiqBw8Bi1bMkDh"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"klmUIackbljiy"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"description"}}]},"finish_reason":null}],"usage":null,"obfuscation":"62x1gwB"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"DHl8TIq3cmdG4"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"g"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Run"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Xerjxz3NKOK4yVY"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" the"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1NrH9SEB3P0odU"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/m"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"p3C26iB8WUCEY4p"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ult"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T1pugjo0yoW79om"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" program"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5BD3Pj4Squ"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ied"}}]},"finish_reason":null}],"usage":null,"obfuscation":"B412ahFdNadpuIn"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" to"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BkbQpN1oOHdxEtm"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"it"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" verify"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nGfybm9Pkqy"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rktwoRNCemTw9"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" output"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rOZnce0nhSr"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bUiMNBVTf8DqL"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" matches"}}]},"finish_reason":null}],"usage":null,"obfuscation":"eHpm8lJxM6"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nD2n2ftIfIvCk12"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" expectations"}}]},"finish_reason":null}],"usage":null,"obfuscation":"67Jiu"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zi7nQJOKwTQEBOB"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XgC7VUdM04wdYQy"} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNqlPoG29jxLwuoTgBGbRzMUkIO","object":"chat.completion.chunk","created":1761663282,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":10113,"completion_tokens":37,"total_tokens":10150,"prompt_tokens_details":{"cached_tokens":9856,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"XN"} + data: {"id":"chatcmpl-CVgwi44UF7u9d8BgHdCDJpGmEqQcV","object":"chat.completion.chunk","created":1761669288,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9863,"completion_tokens":51,"total_tokens":9914,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"QWG3"} data: [DONE] @@ -607,15 +611,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.360623791s + duration: 2.6867995s - id: 5 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46281 + content_length: 45045 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\",\"tool_call_id\":\"call_JlXpqBYdqtlfFG5JZWSEbRoo\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_p6iPAE1CcqORgJwTF1Mh2Dgd\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_p6iPAE1CcqORgJwTF1Mh2Dgd\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_gk3SQodQcQC8MDUFYDdkM7yt\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"edits\\\":[{\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"},{\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"}]}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 2 edits to file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"call_gk3SQodQcQC8MDUFYDdkM7yt\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_l8oMEvpMDIkvfbAq584WE9ir\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\t// Greeting\\n 7|\\tfmt.Println(\\\"Hello, Crush!\\\")\\n 8|}\\n\\n\",\"tool_call_id\":\"call_l8oMEvpMDIkvfbAq584WE9ir\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_1EZWwA7F6vCnLSswkrzCH0G9\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_YF7ugjL9nlAcYgcMrSmtonJ4\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"limit\\\":2000}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_YF7ugjL9nlAcYgcMrSmtonJ4\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_7awYAPC9lIeDqVpsfkNcGmiq\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"edits\\\":[{\\\"old_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\\"new_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\"},{\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"}]}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 2 edits to file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"call_7awYAPC9lIeDqVpsfkNcGmiq\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_1bRloNLG6Jgx4JsfmNpQYRdd\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"go run /tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool/main.go\\\",\\\"description\\\":\\\"Run the Go program to verify output matches expectations\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Hello, Crush!\\n\\n\\n/tmp/crush-test/TestCoderAgent/openai-gpt-5/multiedit_tool\",\"tool_call_id\":\"call_1bRloNLG6Jgx4JsfmNpQYRdd\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -631,13 +635,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNtMFbX4cARf3xcr574GO7qSdd2","object":"chat.completion.chunk","created":1761663285,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"VKW7nvRB4s"} + data: {"id":"chatcmpl-CVgwl49Ko7fl86SvnMp94z1EhEID9","object":"chat.completion.chunk","created":1761669291,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"01slanMHjV"} - data: {"id":"chatcmpl-CVfNtMFbX4cARf3xcr574GO7qSdd2","object":"chat.completion.chunk","created":1761663285,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"MKpjWixw"} + data: {"id":"chatcmpl-CVgwl49Ko7fl86SvnMp94z1EhEID9","object":"chat.completion.chunk","created":1761669291,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"BPWTQvXW"} - data: {"id":"chatcmpl-CVfNtMFbX4cARf3xcr574GO7qSdd2","object":"chat.completion.chunk","created":1761663285,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"W8obSE"} + data: {"id":"chatcmpl-CVgwl49Ko7fl86SvnMp94z1EhEID9","object":"chat.completion.chunk","created":1761669291,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"i6PgKa"} - data: {"id":"chatcmpl-CVfNtMFbX4cARf3xcr574GO7qSdd2","object":"chat.completion.chunk","created":1761663285,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":10220,"completion_tokens":4,"total_tokens":10224,"prompt_tokens_details":{"cached_tokens":9856,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"hrL"} + data: {"id":"chatcmpl-CVgwl49Ko7fl86SvnMp94z1EhEID9","object":"chat.completion.chunk","created":1761669291,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9952,"completion_tokens":4,"total_tokens":9956,"prompt_tokens_details":{"cached_tokens":9856,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"E5FI0"} data: [DONE] @@ -646,4 +650,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.797561042s + duration: 1.557571458s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/parallel_tool_calls.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/parallel_tool_calls.yaml index b9c7f635f4819509a3cf6481c17d8017186e9fc5..2a9c2500d19fcfd825c00f19ca50a0a85c02fba1 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/parallel_tool_calls.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/parallel_tool_calls.yaml @@ -24,25 +24,23 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"eDZ2GmsL0XZzjl"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"YcSHCvRzXF9znf"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Parallel"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"0zbJiurp"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Parallel"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"l8ktWs1G"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Execution"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"pmCgRZ"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" File"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"8wL5OE02PmV"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" of"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"71IAMyDPt4LQD"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Search"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"W6Xqjc0aC"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Glob"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"uHKoUxHcjHH"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" and"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"TP5ta9bRE4aF"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" and"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"DKep9MJQK5dr"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Directory"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"cQoN7H"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" LS"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"RxV0Npzrs9qH4"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Listing"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"QKTVuu8Q"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Commands"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"UYQTC5Y"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"oRiKqwdGEs"} - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"XaRIJd8GS0"} - - data: {"id":"chatcmpl-CVfOXTNvCUoluDkZN5qO600nXHLey","object":"chat.completion.chunk","created":1761663325,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":154,"completion_tokens":7,"total_tokens":161,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"cbGRSghi3E0d9n"} + data: {"id":"chatcmpl-CVgxK3PcrBEGR6zkviVBx6ApLQZUP","object":"chat.completion.chunk","created":1761669326,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":154,"completion_tokens":6,"total_tokens":160,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"rHQpPCIUqCmHmd"} data: [DONE] @@ -51,15 +49,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 527.81725ms + duration: 906.014125ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44238 + content_length: 43021 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -75,27 +73,37 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null},"finish_reason":null}],"obfuscation":"m7u9gSj"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null},"finish_reason":null}],"obfuscation":"eIwWbKg"} + + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"id":"call_Tn8QdGSFthpS1gTTn9bIOIlj","type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null}],"obfuscation":"1eu3aWTW3odCw"} + + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"pa"}}]},"finish_reason":null}],"obfuscation":"qE5vuNlEbBfkF"} + + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ttern"}}]},"finish_reason":null}],"obfuscation":"Ng20X87Idxh9H"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"id":"call_KLsrLCnMM6u74bT9W7WCIgaU","type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null}],"obfuscation":"EbnzqbgYPn2zf"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\": \"**"}}]},"finish_reason":null}],"obfuscation":"1XbQGlX8Ig"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\"pa"}}]},"finish_reason":null}],"obfuscation":"17YfLPTJUxt2y"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/*.g"}}]},"finish_reason":null}],"obfuscation":"Q07yfhyMkNCe1p"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ttern"}}]},"finish_reason":null}],"obfuscation":"uEaC428qeRoug"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"o\"}"}}]},"finish_reason":null}],"obfuscation":"zAh7DZGQaifTMk"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\": \"**"}}]},"finish_reason":null}],"obfuscation":"Uwi6huVU1z"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"id":"call_Ggll4HofZyjxiWocnKNdGmZ9","type":"function","function":{"name":"ls","arguments":""}}]},"finish_reason":null}],"obfuscation":"itArjJ6ecYk39Wb"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/*.g"}}]},"finish_reason":null}],"obfuscation":"qsVqADeeBvI5xF"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"{\"pa"}}]},"finish_reason":null}],"obfuscation":"6oXBylPjVJbCg"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"o\"}"}}]},"finish_reason":null}],"obfuscation":"dw2h11MZ5WPkQW"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"th\": "}}]},"finish_reason":null}],"obfuscation":"EvJx8owDtTB6"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"id":"call_OT0z1UrQtlHXxf4NV8d5ceZ1","type":"function","function":{"name":"ls","arguments":""}}]},"finish_reason":null}],"obfuscation":"DtGabOoBmLeoQmO"} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"\".\", \""}}]},"finish_reason":null}],"obfuscation":"45meCEygI"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"{}"}}]},"finish_reason":null}],"obfuscation":""} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"dept"}}]},"finish_reason":null}],"obfuscation":"56HyS74HmzgfTL"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"h\": 1"}}]},"finish_reason":null}],"obfuscation":"ElvNaAhqeVLo"} - data: {"id":"chatcmpl-CVfOYjxGJ85OgTdnjRdJKEpnxBtNc","object":"chat.completion.chunk","created":1761663326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9760,"completion_tokens":244,"total_tokens":10004,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":192,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"usage":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":1,"function":{"arguments":"}"}}]},"finish_reason":null}],"obfuscation":"H"} + + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgxKYhn31gULwBdRwqk0miEdhYTE","object":"chat.completion.chunk","created":1761669326,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9509,"completion_tokens":570,"total_tokens":10079,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":512,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} data: [DONE] @@ -104,15 +112,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 6.797483917s + duration: 11.018835917s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44816 + content_length: 43628 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_KLsrLCnMM6u74bT9W7WCIgaU\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"call_OT0z1UrQtlHXxf4NV8d5ceZ1\",\"function\":{\"arguments\":\"{}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls/main.go\",\"tool_call_id\":\"call_KLsrLCnMM6u74bT9W7WCIgaU\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_OT0z1UrQtlHXxf4NV8d5ceZ1\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_Tn8QdGSFthpS1gTTn9bIOIlj\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"call_Ggll4HofZyjxiWocnKNdGmZ9\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\".\\\", \\\"depth\\\": 1}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls/main.go\",\"tool_call_id\":\"call_Tn8QdGSFthpS1gTTn9bIOIlj\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_Ggll4HofZyjxiWocnKNdGmZ9\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -128,65 +136,113 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"VU9wD8UXaE"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"BhwWBpy0ob"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"6YSgYYKJr1U"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" /"},"finish_reason":null}],"usage":null,"obfuscation":"DIroOL6lE4"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"1ok2XEui6"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"uHFgwcVVT"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"MFgoriV3m"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"aUaWdtx"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"7UvzyWL"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"rWU1nTn"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"6W69ND4"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"gySkOk6"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"30NneugqD1"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"DWQr2wVCAs"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"N0YnGcEBIv"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"ZIf9yECfb4H"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"LXkL8ySEcg8"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"DCAtkXqW3LV"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"parallel"},"finish_reason":null}],"usage":null,"obfuscation":"ujH6"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"83b1ZkI"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_calls"},"finish_reason":null}],"usage":null,"obfuscation":"TcK6rF"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"0a2mkot"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"68b1AIEZA"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"VKx0fna4dU"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"aEGT0pv80Us"} + + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" /"},"finish_reason":null}],"usage":null,"obfuscation":"thPSgScqD2"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"I4bfvdisUAi"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"CBy03uPLI"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" /"},"finish_reason":null}],"usage":null,"obfuscation":"pHk5PZhtLS"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"Q4GznvGoT"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"tmp"},"finish_reason":null}],"usage":null,"obfuscation":"OG2cSaX5C"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"dEoRBMMMD"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/cr"},"finish_reason":null}],"usage":null,"obfuscation":"PsgvlTxP7"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"jsUWzX6"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ush"},"finish_reason":null}],"usage":null,"obfuscation":"azGyI2lDH"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"vGSJ6bX"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-test"},"finish_reason":null}],"usage":null,"obfuscation":"t8bSyzp"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"xxWjncC"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/Test"},"finish_reason":null}],"usage":null,"obfuscation":"7ej6Ukz"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"4vCCH7c"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Coder"},"finish_reason":null}],"usage":null,"obfuscation":"yHd8E23"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"eo6clNs"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Agent"},"finish_reason":null}],"usage":null,"obfuscation":"NJfQ1hb"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"KuwxcFCRpC"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/open"},"finish_reason":null}],"usage":null,"obfuscation":"7b2QtHC"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"30LFqoO9Js"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ai"},"finish_reason":null}],"usage":null,"obfuscation":"ciiaqLM1sa"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"X65fMP8rC9"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-g"},"finish_reason":null}],"usage":null,"obfuscation":"wO3l4I91ap"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"qqdK02Ja1mD"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"pt"},"finish_reason":null}],"usage":null,"obfuscation":"22o6Q4st5T"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"ctw7ktbgAHo"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"am3A78PQPcH"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"e1Y8sp4tJKu"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"5"},"finish_reason":null}],"usage":null,"obfuscation":"um7cCn2SXC6"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"parallel"},"finish_reason":null}],"usage":null,"obfuscation":"1CvB"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"U6RYu3qkS5J"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"OFBmHSu"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"parallel"},"finish_reason":null}],"usage":null,"obfuscation":"yzxH"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_calls"},"finish_reason":null}],"usage":null,"obfuscation":"cX2zwx"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_tool"},"finish_reason":null}],"usage":null,"obfuscation":"7T30AxA"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/\n"},"finish_reason":null}],"usage":null,"obfuscation":"XhEAbVdBW"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"_calls"},"finish_reason":null}],"usage":null,"obfuscation":"EeKAzY"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}],"usage":null,"obfuscation":"suimTlKDpF7"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"OSI6gp6"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" -"},"finish_reason":null}],"usage":null,"obfuscation":"WYxBzt0Kka"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"o9ng7FEn6"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" go"},"finish_reason":null}],"usage":null,"obfuscation":"A9ZgCYvug"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"FZHFUhgFVb"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}],"usage":null,"obfuscation":"Gyjqtk3J"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}],"usage":null,"obfuscation":"2vC6T7pTGu"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"FqKFHvZBWT"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".mod"},"finish_reason":null}],"usage":null,"obfuscation":"saSnI65z"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}],"usage":null,"obfuscation":"fzNDLrcLwpu"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":","},"finish_reason":null}],"usage":null,"obfuscation":"RgYV2F2R5VT"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" -"},"finish_reason":null}],"usage":null,"obfuscation":"W18YHBbqO0"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}],"usage":null,"obfuscation":"4I0uMEI"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" main"},"finish_reason":null}],"usage":null,"obfuscation":"89BAPl0"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"qbh30LBvu"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"epm77IGTL"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"l700Vh"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"KhnuSW"} - data: {"id":"chatcmpl-CVfOejQeCaeQHLCgJxeHpoSeMqDTx","object":"chat.completion.chunk","created":1761663332,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9875,"completion_tokens":30,"total_tokens":9905,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"cPJk"} + data: {"id":"chatcmpl-CVgxWrTbkLJWnfCjyp8qf6bP3bdJ8","object":"chat.completion.chunk","created":1761669338,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9630,"completion_tokens":54,"total_tokens":9684,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"lNcY"} data: [DONE] @@ -195,4 +251,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.553797542s + duration: 1.311035167s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/read_a_file.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/read_a_file.yaml index b836f2997d673c751e9a3f3ac1c834be66f93db9..4a99486822e36ea6cb527b71eefa849d45690d26 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/read_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/read_a_file.yaml @@ -24,17 +24,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMDiln9OYwJuTwhLVxkTnuUAInN","object":"chat.completion.chunk","created":1761663181,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"WOL9laQ1Ce56Jh"} + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"CfUidPnEqWPXNf"} - data: {"id":"chatcmpl-CVfMDiln9OYwJuTwhLVxkTnuUAInN","object":"chat.completion.chunk","created":1761663181,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":"Reading"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Vrh4IHpun"} + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":"Reading"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"WRVjSOd3d"} - data: {"id":"chatcmpl-CVfMDiln9OYwJuTwhLVxkTnuUAInN","object":"chat.completion.chunk","created":1761663181,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" Go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Pwh05Ly8CtxiP"} + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" Go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"nEYDzTwVQC08G"} - data: {"id":"chatcmpl-CVfMDiln9OYwJuTwhLVxkTnuUAInN","object":"chat.completion.chunk","created":1761663181,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" Modules"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"hPUiFBe7"} + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" Mod"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"LoeNmRdBi1tU"} - data: {"id":"chatcmpl-CVfMDiln9OYwJuTwhLVxkTnuUAInN","object":"chat.completion.chunk","created":1761663181,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"hvtqQSjs4i"} + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{"content":" Files"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"G7qPCFFe3p"} - data: {"id":"chatcmpl-CVfMDiln9OYwJuTwhLVxkTnuUAInN","object":"chat.completion.chunk","created":1761663181,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[],"usage":{"prompt_tokens":129,"completion_tokens":3,"total_tokens":132,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"HMEDXy9sgeWpNQ"} + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"C51VEgb3KU"} + + data: {"id":"chatcmpl-CVgu1i5F96gR2MvpLO1VSeMEToyAS","object":"chat.completion.chunk","created":1761669121,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_a788c5aef0","choices":[],"usage":{"prompt_tokens":129,"completion_tokens":4,"total_tokens":133,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"ywVNSMw56Eg4I6"} data: [DONE] @@ -43,15 +45,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 695.326541ms + duration: 1.032028958s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44111 + content_length: 42894 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -67,61 +69,63 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_4xE2NF0D4dTTbRcPHdbmr84w","type":"function","function":{"name":"ls","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"sqXguISTvNnmqq"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_NBKeQUSLmayh4GY17GEkhxjx","type":"function","function":{"name":"glob","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"HDNKs4Mwp16l"} + + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"pn4idLeRaHXnGrK"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"7eUOr7tM7Xbymgh"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pattern"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rnNQRlIgRUR"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Qq5mDlC3bXJZJc"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"NL9X5C5rd2Ys5"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OgahnzGO3hfcq1L"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uZbpGkldiz1V6WO"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".mod"}}]},"finish_reason":null}],"usage":null,"obfuscation":"EnAjHerzoHIoiv"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fy50QXAuwjoS4Nv"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"yTj81c04DcKzJ"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"93BSs3GVBJlmB8D"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jS1gymP1jiAqgR"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"PHb8y2jPpsLtsMK"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Vs4bEXliPWz2NCi"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kKwfjFsZden5P"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M8rOjQyc7QLCyeB"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IsNVlRQE0uS5y"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hSTR3pg7Gu0Ky5O"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Vz0DtgsrLtIC8"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"JNPoeXldUQXMkcw"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Xz42OBsQ3UJ1X"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"slhQI45wskSCugs"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nmGCqm6MrT2hk"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YRdrnNG2NnzPk"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NGDtrwURo0Vc5"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"VqMjz4MJ871Nt"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WTzwiepnjXRmS"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"s"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uP48U8jWBR3XA"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"E"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/read"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OihKAakxSjtRK"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D1YYaKhNoJKd6"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"tqSBGyPF5nci8"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"W"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"depth"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Hl9vizXhvKg5O"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/read"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jXBPLwkFgsGlF"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HMtetoqF8Y1NHu7"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"2"}}]},"finish_reason":null}],"usage":null,"obfuscation":"i"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QiGs2H0QM4xDV"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"l"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ITWZ1QsgvWG38KJ"} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMD1QIB4oTZrYSI8MHbVHSDyYYq","object":"chat.completion.chunk","created":1761663181,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9734,"completion_tokens":171,"total_tokens":9905,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":128,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"K"} + data: {"id":"chatcmpl-CVgu2PIn59QvFu2lO8Cy6iD4PA3t7","object":"chat.completion.chunk","created":1761669122,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9483,"completion_tokens":299,"total_tokens":9782,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":256,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"t"} data: [DONE] @@ -130,15 +134,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.896481833s + duration: 5.787486416s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44489 + content_length: 43260 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_4xE2NF0D4dTTbRcPHdbmr84w\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\\",\\\"depth\\\":2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_4xE2NF0D4dTTbRcPHdbmr84w\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_NBKeQUSLmayh4GY17GEkhxjx\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"go.mod\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file/go.mod\",\"tool_call_id\":\"call_NBKeQUSLmayh4GY17GEkhxjx\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -154,69 +158,59 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_54RSa4IlGvh0PbruYbWJwMwp","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"YuCAQQAEWIgf"} - - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"fZmlHWiB3zFp2zY"} - - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1pxQSwwVv0tjIS"} - - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wdyMXOdmSVz3O"} - - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kmRKvkV5x9QTy0R"} - - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Pkc5FWVU7vpQvJD"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_T9Y7DWidvJVzEKFL6nW30WGz","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"SdPtqiow5bWC"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nBb2rQ45eqWmM28"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"HzbRhVYOXedjbvG"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0YznXh3Zx1ER89s"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nFzVtfIyhedJg7"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"EpoOLxN9b3F5zM4"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sq34OxJGhB1qW"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"mQ9wKWAAoGuJU"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NrB6oOiNMmctngt"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wx7igy2xRGH6N"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TCasVHKEru7yB9E"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cnvLqaDVaE4Dz"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Y07VsaMcQYR2t8K"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"B9RO8X79u4S6D"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AL85w4BzUtB1LMe"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"acNl4KhnyK6zs"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iUatY5nRVNPurYd"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QXpQbamIqsXRJ"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AEO5UE8fRLMYV"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8td7sCZcuQlVr"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"P"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"b6X281P1YVQ9h"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"m"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hI2sTGt1q4ouS"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/read"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ymiRYNCpfX5NG"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LkLgTllmZYoXY"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xIhICIWfYjRu4D3"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"m"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".mod"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wECQZy4FCy3xQO"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"U"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"VfKnnORcUqTcn"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/read"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kqkgTHthFlUc9"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"limit"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ogUvP4tUatLFd"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"I34QmsrJ7vwEgpw"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"btGru9zcUAMBI"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"200"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q58vARlzg74ISxw"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"k0VPPBTUvEkeAvB"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"0"}}]},"finish_reason":null}],"usage":null,"obfuscation":"R"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".mod"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MLUir18h3jZ5YV"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"v"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sl7HNuW8fhJvuHU"} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMH6NCVWJeIBI4UsYMQu2jBLGN1","object":"chat.completion.chunk","created":1761663185,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9811,"completion_tokens":40,"total_tokens":9851,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Pere"} + data: {"id":"chatcmpl-CVgu7OsDz6hicem9Z5t1lD7lOl8oA","object":"chat.completion.chunk","created":1761669127,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9548,"completion_tokens":35,"total_tokens":9583,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"nBce"} data: [DONE] @@ -225,15 +219,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 4.818397166s + duration: 1.892169834s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44878 + content_length: 43634 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_4xE2NF0D4dTTbRcPHdbmr84w\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\\",\\\"depth\\\":2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_4xE2NF0D4dTTbRcPHdbmr84w\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_54RSa4IlGvh0PbruYbWJwMwp\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file/go.mod\\\",\\\"limit\\\":2000}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\n\",\"tool_call_id\":\"call_54RSa4IlGvh0PbruYbWJwMwp\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_NBKeQUSLmayh4GY17GEkhxjx\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"go.mod\\\",\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file/go.mod\",\"tool_call_id\":\"call_NBKeQUSLmayh4GY17GEkhxjx\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_T9Y7DWidvJVzEKFL6nW30WGz\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\n\",\"tool_call_id\":\"call_T9Y7DWidvJVzEKFL6nW30WGz\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -249,33 +243,33 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"QkoFIHp2Wv"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"7Tbzg5ScWS"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"module"},"finish_reason":null}],"usage":null,"obfuscation":"umKBvx"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"module"},"finish_reason":null}],"usage":null,"obfuscation":"qWcCM7"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}],"usage":null,"obfuscation":"aPYY"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" example"},"finish_reason":null}],"usage":null,"obfuscation":"A94T"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"nju7XOma"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"eIWlL0Lr"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/test"},"finish_reason":null}],"usage":null,"obfuscation":"bzw0dgh"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/test"},"finish_reason":null}],"usage":null,"obfuscation":"lbzzAdv"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"project"},"finish_reason":null}],"usage":null,"obfuscation":"JC3z7"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"project"},"finish_reason":null}],"usage":null,"obfuscation":"VNCRV"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"gGN8jMzLuY"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n\n"},"finish_reason":null}],"usage":null,"obfuscation":"cdUJOuQC"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}],"usage":null,"obfuscation":"sFx3yogj6V"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"go"},"finish_reason":null}],"usage":null,"obfuscation":"jYgJq71AY0"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}],"usage":null,"obfuscation":"v1mgVnjx8a6"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}],"usage":null,"obfuscation":"WObvKw5L6FS"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}],"usage":null,"obfuscation":"pjvABWN478a"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"1"},"finish_reason":null}],"usage":null,"obfuscation":"8VOUe88l4JI"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"."},"finish_reason":null}],"usage":null,"obfuscation":"XEmO3wevBJZ"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"."},"finish_reason":null}],"usage":null,"obfuscation":"2AjAP6315qU"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"23"},"finish_reason":null}],"usage":null,"obfuscation":"i5JJOPKh7O"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"23"},"finish_reason":null}],"usage":null,"obfuscation":"ddbjtmxAfU"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"sapCl1"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"TC5gWd"} - data: {"id":"chatcmpl-CVfMMJPcfio38PsvlbrefxLKoWWfu","object":"chat.completion.chunk","created":1761663190,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9891,"completion_tokens":14,"total_tokens":9905,"prompt_tokens_details":{"cached_tokens":9728,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"ocXK"} + data: {"id":"chatcmpl-CVguBnwG6kJ5xSQTqUlOjsv8KuGTG","object":"chat.completion.chunk","created":1761669131,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9623,"completion_tokens":14,"total_tokens":9637,"prompt_tokens_details":{"cached_tokens":9472,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"fBOi"} data: [DONE] @@ -284,4 +278,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.339726042s + duration: 1.924837166s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/simple_test.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/simple_test.yaml index 6addfc4b0c239f6e4437c77c7eaa926098ba0949..d70c7bcab99a720a8e4e05beff9dbf43fb120a7e 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/simple_test.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/simple_test.yaml @@ -24,15 +24,15 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfM8qm9phmikdosW4EUrnhhD9GDV","object":"chat.completion.chunk","created":1761663176,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Ikc5EPxVah532d"} + data: {"id":"chatcmpl-CVgtxoHQz9ZhnWjUqbZxqc1A5Zf5z","object":"chat.completion.chunk","created":1761669117,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"1qmAaiIaFnmqqD"} - data: {"id":"chatcmpl-CVfM8qm9phmikdosW4EUrnhhD9GDV","object":"chat.completion.chunk","created":1761663176,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Greeting"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"d9HjQK6a"} + data: {"id":"chatcmpl-CVgtxoHQz9ZhnWjUqbZxqc1A5Zf5z","object":"chat.completion.chunk","created":1761669117,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Greetings"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Wqcbg9W"} - data: {"id":"chatcmpl-CVfM8qm9phmikdosW4EUrnhhD9GDV","object":"chat.completion.chunk","created":1761663176,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Inquiry"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"9XM31cf0"} + data: {"id":"chatcmpl-CVgtxoHQz9ZhnWjUqbZxqc1A5Zf5z","object":"chat.completion.chunk","created":1761669117,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Inquiry"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"snTmXVg1"} - data: {"id":"chatcmpl-CVfM8qm9phmikdosW4EUrnhhD9GDV","object":"chat.completion.chunk","created":1761663176,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"vndlpXzHkm"} + data: {"id":"chatcmpl-CVgtxoHQz9ZhnWjUqbZxqc1A5Zf5z","object":"chat.completion.chunk","created":1761669117,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"ntogY4rTB0"} - data: {"id":"chatcmpl-CVfM8qm9phmikdosW4EUrnhhD9GDV","object":"chat.completion.chunk","created":1761663176,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":126,"completion_tokens":2,"total_tokens":128,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"2SDl90JyBGlr6L"} + data: {"id":"chatcmpl-CVgtxoHQz9ZhnWjUqbZxqc1A5Zf5z","object":"chat.completion.chunk","created":1761669117,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":126,"completion_tokens":2,"total_tokens":128,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"zlR2J2wkcUzURC"} data: [DONE] @@ -41,15 +41,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 722.778417ms + duration: 1.672707833s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44101 + content_length: 42884 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Hello\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Hello\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -65,13 +65,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfM90feAsox6lcGe5oW1R3CdDH55","object":"chat.completion.chunk","created":1761663177,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"eYOwlkp1HO"} + data: {"id":"chatcmpl-CVgtxASLdlIkKGm4lpQFSYVC0jZ0d","object":"chat.completion.chunk","created":1761669117,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"8CWM8Eqgte"} - data: {"id":"chatcmpl-CVfM90feAsox6lcGe5oW1R3CdDH55","object":"chat.completion.chunk","created":1761663177,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Hi"},"finish_reason":null}],"usage":null,"obfuscation":"k4Fee3TpAY"} + data: {"id":"chatcmpl-CVgtxASLdlIkKGm4lpQFSYVC0jZ0d","object":"chat.completion.chunk","created":1761669117,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Hi"},"finish_reason":null}],"usage":null,"obfuscation":"3WHdeJu6Oq"} - data: {"id":"chatcmpl-CVfM90feAsox6lcGe5oW1R3CdDH55","object":"chat.completion.chunk","created":1761663177,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"7a083Y"} + data: {"id":"chatcmpl-CVgtxASLdlIkKGm4lpQFSYVC0jZ0d","object":"chat.completion.chunk","created":1761669117,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"0NTzQ0"} - data: {"id":"chatcmpl-CVfM90feAsox6lcGe5oW1R3CdDH55","object":"chat.completion.chunk","created":1761663177,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9730,"completion_tokens":74,"total_tokens":9804,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":64,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"W9j"} + data: {"id":"chatcmpl-CVgtxASLdlIkKGm4lpQFSYVC0jZ0d","object":"chat.completion.chunk","created":1761669117,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9479,"completion_tokens":10,"total_tokens":9489,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"CDaeXSj"} data: [DONE] @@ -80,4 +80,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 3.682005167s + duration: 3.76380425s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/sourcegraph_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/sourcegraph_tool.yaml index eb778515d79ec6edb398b9a660270793fee8c825..9301cd56a987f9227488e18cbe3cf78f9508f213 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/sourcegraph_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/sourcegraph_tool.yaml @@ -24,35 +24,31 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ePizzDvdvS6CrU"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"JvsGFizrmrNTcq"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Search"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"UwRRLOgUg4"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Searching"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"tzoiD2V"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" '"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"2gkuMdNnyXSn03"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" for"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"D7k0fm84144e"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"func"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"uOCSSLoqRpiQ"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" '"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"9Er4045AMW4UT5"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" main"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"khUNjcNbx1I"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"func"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"51fj1NW4oCfl"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"'"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"P8VlVgH6yWwxlcr"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" main"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"8zhWtFF5sGT"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"y44gM4c1Qtiyj"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"'"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"wuTukmBsaUvjQi9"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"PK8qJRLFizP9X"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"NSBb902gHLV3T"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Re"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"QHUbxx9TytDyN"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"bJAKtPKqd1E37"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"positories"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"rvA3qD"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Re"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"bdXqOkzot7t0d"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"Of1CcuI7w8x"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"positories"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"g5RE3D"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Source"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"oz35XxN1g"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"0SGP5GM8J0"} - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"graph"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"0LLvtRf8STU"} - - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"HRk7trwLHM"} - - data: {"id":"chatcmpl-CVfNv6CiDLShGMZN0Cgfl1cVqrLis","object":"chat.completion.chunk","created":1761663287,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":138,"completion_tokens":12,"total_tokens":150,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"MwkK4SmMQ0Wo1"} + data: {"id":"chatcmpl-CVgwnkKl1lPaWcSnSG1bqNQlwSJKb","object":"chat.completion.chunk","created":1761669293,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":138,"completion_tokens":10,"total_tokens":148,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"IbSVuiEydGINz"} data: [DONE] @@ -61,15 +57,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 496.019625ms + duration: 784.394792ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44161 + content_length: 42944 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -85,55 +81,65 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_WLV6Tr9aRugjPKQxpqRJvzIW","type":"function","function":{"name":"sourcegraph","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"NYUbO"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_usUSsyqHEqfNtcKQR2sSKbZn","type":"function","function":{"name":"sourcegraph","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"h5z5T"} + + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"W33MvkYxQB1RSQq"} + + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"query"}}]},"finish_reason":null}],"usage":null,"obfuscation":"DjpYQ8xRdydzk"} + + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"d0PiRCHlmTyHx"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"w7YzyHH9jPKG5aX"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lang"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZF9GRNLdzUSJJQ"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"query"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iZ6uYREUwHM0Z"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"v"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"8Lqekn5VEC94m"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"lang"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jsKGsZ6rVFKONt"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"89MQKqyq8S13r"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"I"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9gdbsOeLhEbOZY"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kMrId3vbV96r5"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"AUilwTsLFL63H"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"dVjy5shIjUTBM1"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bIIWDtIm6LfyY9"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"j7pCFBhvO3YCa"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TCMQH1W7wUSbS"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":":\\\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QVoGMzFLLtBcs"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\"\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"ajpIcSaUPxB"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"."}}]},"finish_reason":null}],"usage":null,"obfuscation":"O"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"GtastoCKainW5lu"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"count"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3n6tXz8C6praa"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"$\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"BAu91f4P7ROnLt"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BwFf6v8xCWVXpzD"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"u7W866UE9mdvqXA"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"z"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"count"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uKHqdjbAAVuP3"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"10"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cepCzN95s6x5870"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"l"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"N"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"fIkdMCNY31ribjF"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"10"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"timeout"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sxWrqikkvYf"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"0"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MdupbA8FscIqp3D"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"4BDGplBKviNv69b"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"5"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"timeout"}}]},"finish_reason":null}],"usage":null,"obfuscation":"30WxOnc5gXz"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"60"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"VhGCSL3sRGbcKPu"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1"} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"J"} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"60"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfNvEKtbkXZDPh1KFVWtFlWJhkdj","object":"chat.completion.chunk","created":1761663287,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9743,"completion_tokens":424,"total_tokens":10167,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":384,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A"} + + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + + data: {"id":"chatcmpl-CVgwnE4qNYQ6UhilNbnhclso5A2yc","object":"chat.completion.chunk","created":1761669293,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9492,"completion_tokens":237,"total_tokens":9729,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":192,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"d"} data: [DONE] @@ -142,15 +148,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 9.27407525s + duration: 5.417793625s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 428 + content_length: 440 host: sourcegraph.com - body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"lang:go \"func main\""}}' + body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"lang:go \"func main\" file:\\.go$"}}' headers: Content-Type: - application/json @@ -170,15 +176,15 @@ interactions: - application/json status: 200 OK code: 200 - duration: 1.339548334s + duration: 1.382094541s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 65340 + content_length: 64137 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_WLV6Tr9aRugjPKQxpqRJvzIW\",\"function\":{\"arguments\":\"{\\\"query\\\":\\\"lang:go \\\\\\\"func main\\\\\\\"\\\", \\\"count\\\": 10, \\\"timeout\\\": 60}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/wx-chevalier/Go-Notes/99~参考资料/2017~A Huge Number of Go Examples/codes/first/first/main.go\\n\\nURL: /github.com/wx-chevalier/Go-Notes/-/blob/99~%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/2017~A%20Huge%20Number%20of%20Go%20Examples/codes/first/first/main.go\\n\\n```\\n6| /*\\n7| This is a multi-line comment.\\n8| \\n9| import keyword makes another package available\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| // \\\"func main\\\" is special.\\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| func main() {\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// godoc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n```\\n\\n```\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| // func main creates a starting point for Go\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// godoc -src fmt Println\\n```\\n\\n```\\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// godoc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n35| \\t// Exported = First Letter is uppercase\\n36| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n37| \\n38| \\t// Go cannot call Println function by itself.\\n39| \\t// It only calls `func main` automatically.\\n40| \\t// It only calls `func main` automatically.\\n41| \\n42| \\t// -----\\n43| \\n44| \\t// Go supports Unicode characters in string literals\\n45| \\t// And also in source-code: KÖSTEBEK!\\n46| \\n47| \\t// EXERCISE: Remove the comments from below --> //\\n48| \\t// fmt.Println(\\\"Merhaba Köstebek!\\\")\\n49| \\n```\\n\\n## Result 2: github.com/wx-chevalier/Go-Notes/99~参考资料/2017~A Huge Number of Go Examples/codes/02-write-your-first-program/main.go\\n\\nURL: /github.com/wx-chevalier/Go-Notes/-/blob/99~%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/2017~A%20Huge%20Number%20of%20Go%20Examples/codes/02-write-your-first-program/main.go\\n\\n```\\n6| /*\\n7| This is a multi-line comment.\\n8| \\n9| import keyword makes another package available\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| // \\\"func main\\\" is special.\\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| func main() {\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// go doc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n```\\n\\n```\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| // func main creates a starting point for Go\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// go doc -src fmt Println\\n```\\n\\n```\\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// go doc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n35| \\t// Exported = First Letter is uppercase\\n36| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n37| \\n38| \\t// Go cannot call Println function by itself.\\n39| \\t// It only calls `func main` automatically.\\n40| \\t// It only calls `func main` automatically.\\n41| \\n42| \\t// -----\\n43| \\n44| \\t// Go supports Unicode characters in string literals\\n45| \\t// And also in source-code: KÖSTEBEK!\\n46| \\t//\\n47| \\t// Because: Literal ~= Source Code\\n48| \\n49| \\t// EXERCISE: Remove the comments from below --> //\\n```\\n\\n## Result 3: github.com/Go-Learning-Land/Go-Learning-Archive/GolangBasics/learngo/translation/spanish/02-tu-primer-programa/main.go\\n\\nURL: /github.com/Go-Learning-Land/Go-Learning-Archive/-/blob/GolangBasics/learngo/translation/spanish/02-tu-primer-programa/main.go\\n\\n```\\n13| /*\\n14| Esto es un comentario multilinea.\\n15| \\n16| la palabra import hace que otro paquete este disponible\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" es especial.\\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Despues de compilar el codigo,\\n31| // Go ejecutara esta funcion primero.\\n32| func main() {\\n33| \\t// despues del: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Despues de compilar el codigo,\\n31| func main() {\\n32| func main() {\\n33| \\t// despues del: import \\\"fmt\\\"\\n34| \\t// La funcion Println del paquete \\\"fmt\\\" estara disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una funcion exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n```\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| // func main crea un punto de inicio para Go\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Despues de compilar el codigo,\\n31| // Go ejecutara esta funcion primero.\\n32| func main() {\\n33| \\t// despues del: import \\\"fmt\\\"\\n34| \\t// La funcion Println del paquete \\\"fmt\\\" estara disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n```\\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una funcion exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n42| \\t// Exportada = Primera letra en mayuscula\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go no puede llamar a la funcion Println por si mismo.\\n46| \\t// Solo llama a la `func main` de forma automatica.\\n47| \\t// Solo llama a la `func main` de forma automatica.\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go soporta caracteres unicode en cadenas de texto literal\\n52| \\t// y tambien en el codigo fuente: KÖSTEBEK!\\n53| \\t//\\n54| \\t// Porque: Literal ~= Codigo Fuente\\n55| }\\n56| \\n```\\n\\n## Result 4: github.com/inancgumus/learngo/translation/chinese/02-编写第一个程序/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/translation/chinese/02-%E7%BC%96%E5%86%99%E7%AC%AC%E4%B8%80%E4%B8%AA%E7%A8%8B%E5%BA%8F/main.go\\n\\n```\\n13| /*\\n14| 这是一个多行注释\\n15| \\n16| import 关键字引入另一个包\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" 是特殊的\\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| func main() {\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n```\\n\\n```\\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n42| \\t// Exported = 首字母大写\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go 本身不能调用 Println 函数\\n46| \\t// Go 只能自动地调用 `func main`\\n47| \\t// Go 只能自动地调用 `func main`\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go 字符串支持 Unicode 字符\\n52| \\t// 源代码同时也支持: KÖSTEBEK!\\n53| \\t//\\n54| \\t// 因为: 字面值 ~= 源代码\\n55| \\n56| \\t// 练习: 删除下行注释的 --> //\\n```\\n\\n```\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| // func main 对于 Go 而言就是开始执行的位置\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n## Result 5: github.com/inancgumus/learngo/first/first/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/first/first/main.go\\n\\n```\\n13| /*\\n14| This is a multi-line comment.\\n15| \\n16| import keyword makes another package available\\n17| for this .go \\\"file\\\".\\n18| \\n19| import \\\"fmt\\\" lets you access fmt package's functionality\\n20| here in this file.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" is special.\\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| // Go runtime will first run this function\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| func main() {\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n34| \\t// Println function of \\\"fmt\\\" package becomes available\\n35| \\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println is just an exported function from\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n```\\n17| for this .go \\\"file\\\".\\n18| \\n19| import \\\"fmt\\\" lets you access fmt package's functionality\\n20| here in this file.\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| // func main creates a starting point for Go\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| // Go runtime will first run this function\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n34| \\t// Println function of \\\"fmt\\\" package becomes available\\n35| \\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n```\\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println is just an exported function from\\n40| \\t// \\\"fmt\\\" package\\n41| \\n42| \\t// Exported = First Letter is uppercase\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go cannot call Println function by itself.\\n46| \\t// It only calls `func main` automatically.\\n47| \\t// It only calls `func main` automatically.\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go supports Unicode characters in string literals\\n52| \\t// And also in source-code: KÖSTEBEK!\\n53| \\n54| \\t// EXERCISE: Remove the comments from below --> //\\n55| \\t// fmt.Println(\\\"Merhaba Köstebek!\\\")\\n56| \\n```\\n\\n## Result 6: github.com/inancgumus/learngo/translation/spanish/02-tu-primer-programa/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/translation/spanish/02-tu-primer-programa/main.go\\n\\n```\\n13| /*\\n14| Esto es un comentario multilinea.\\n15| \\n16| la palabra import hace que otro paquete este disponible\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" es especial.\\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Después de compilar el codigo,\\n31| // Go ejecutará esta función primero.\\n32| func main() {\\n33| \\t// después del: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Después de compilar el codigo,\\n31| func main() {\\n32| func main() {\\n33| \\t// después del: import \\\"fmt\\\"\\n34| \\t// La función Println del paquete \\\"fmt\\\" estará disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una función exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n```\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| // func main crea un punto de inicio para Go\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Después de compilar el codigo,\\n31| // Go ejecutará esta función primero.\\n32| func main() {\\n33| \\t// después del: import \\\"fmt\\\"\\n34| \\t// La función Println del paquete \\\"fmt\\\" estará disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n```\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una función exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n42| \\t// Para poder exportar una función tendrás que escribir el\\n43| \\t// primer caracter del nombre de la función en mayúscula.\\n44| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n45| \\n46| \\t// Go no puede llamar a la función Println por si mismo.\\n47| \\t// Solo llama a la `func main` de forma automatica.\\n48| \\t// Solo llama a la `func main` de forma automatica.\\n49| \\n50| \\t// -----\\n51| \\n52| \\t// Go soporta caracteres unicode en cadenas de texto literal\\n53| \\t// y tambien en el codigo fuente: KÖSTEBEK!\\n54| \\t//\\n55| \\t// Porque: Literal ~= Codigo Fuente\\n56| }\\n57| \\n```\\n\\n## Result 7: github.com/Go-Learning-Land/Go-Learning-Archive/GolangBasics/learngo/translation/chinese/02-编写第一个程序/main.go\\n\\nURL: /github.com/Go-Learning-Land/Go-Learning-Archive/-/blob/GolangBasics/learngo/translation/chinese/02-%E7%BC%96%E5%86%99%E7%AC%AC%E4%B8%80%E4%B8%AA%E7%A8%8B%E5%BA%8F/main.go\\n\\n```\\n13| /*\\n14| 这是一个多行注释\\n15| \\n16| import 关键字引入另一个包\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" 是特殊的\\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| func main() {\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n```\\n\\n```\\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n42| \\t// Exported = 首字母大写\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go 本身不能调用 Println 函数\\n46| \\t// Go 只能自动地调用 `func main`\\n47| \\t// Go 只能自动地调用 `func main`\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go 字符串支持 Unicode 字符\\n52| \\t// 源代码同时也支持: KÖSTEBEK!\\n53| \\t//\\n54| \\t// 因为: 字面值 ~= 源代码\\n55| \\n56| \\t// 练习: 删除下行注释的 --> //\\n```\\n\\n```\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| // func main 对于 Go 而言就是开始执行的位置\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n## Result 8: github.com/inancgumus/learngo/02-write-your-first-program/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/02-write-your-first-program/main.go\\n\\n```\\n13| /*\\n14| This is a multi-line comment.\\n15| \\n16| import keyword makes another package available\\n17| for this .go \\\"file\\\".\\n18| \\n19| import \\\"fmt\\\" lets you access fmt package's functionality\\n20| here in this file.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" is special.\\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| // Go runtime will first run this function\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| func main() {\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n34| \\t// Println function of \\\"fmt\\\" package becomes available\\n35| \\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println is just an exported function from\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n\",\"tool_call_id\":\"call_WLV6Tr9aRugjPKQxpqRJvzIW\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_usUSsyqHEqfNtcKQR2sSKbZn\",\"function\":{\"arguments\":\"{\\\"query\\\":\\\"lang:go \\\\\\\"func main\\\\\\\" file:\\\\\\\\.go$\\\", \\\"count\\\": 10, \\\"timeout\\\": 60}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/wx-chevalier/Go-Notes/99~参考资料/2017~A Huge Number of Go Examples/codes/first/first/main.go\\n\\nURL: /github.com/wx-chevalier/Go-Notes/-/blob/99~%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/2017~A%20Huge%20Number%20of%20Go%20Examples/codes/first/first/main.go\\n\\n```\\n6| /*\\n7| This is a multi-line comment.\\n8| \\n9| import keyword makes another package available\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| // \\\"func main\\\" is special.\\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| func main() {\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// godoc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n```\\n\\n```\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| // func main creates a starting point for Go\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// godoc -src fmt Println\\n```\\n\\n```\\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// godoc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n35| \\t// Exported = First Letter is uppercase\\n36| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n37| \\n38| \\t// Go cannot call Println function by itself.\\n39| \\t// It only calls `func main` automatically.\\n40| \\t// It only calls `func main` automatically.\\n41| \\n42| \\t// -----\\n43| \\n44| \\t// Go supports Unicode characters in string literals\\n45| \\t// And also in source-code: KÖSTEBEK!\\n46| \\n47| \\t// EXERCISE: Remove the comments from below --> //\\n48| \\t// fmt.Println(\\\"Merhaba Köstebek!\\\")\\n49| \\n```\\n\\n## Result 2: github.com/wx-chevalier/Go-Notes/99~参考资料/2017~A Huge Number of Go Examples/codes/02-write-your-first-program/main.go\\n\\nURL: /github.com/wx-chevalier/Go-Notes/-/blob/99~%E5%8F%82%E8%80%83%E8%B5%84%E6%96%99/2017~A%20Huge%20Number%20of%20Go%20Examples/codes/02-write-your-first-program/main.go\\n\\n```\\n6| /*\\n7| This is a multi-line comment.\\n8| \\n9| import keyword makes another package available\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| // \\\"func main\\\" is special.\\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| //\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| func main() {\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// go doc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n```\\n\\n```\\n10| for this .go \\\"file\\\".\\n11| \\n12| import \\\"fmt\\\" lets you access fmt package's functionality\\n13| here in this file.\\n14| */\\n15| import \\\"fmt\\\"\\n16| \\n17| // \\\"func main\\\" is special.\\n18| //\\n19| // Go has to know where to start\\n20| // func main creates a starting point for Go\\n21| // func main creates a starting point for Go\\n22| //\\n23| // After compiling the code,\\n24| // Go runtime will first run this function\\n25| func main() {\\n26| \\t// after: import \\\"fmt\\\"\\n27| \\t// Println function of \\\"fmt\\\" package becomes available\\n28| \\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// go doc -src fmt Println\\n```\\n\\n```\\n29| \\t// Look at what it looks like by typing in the console:\\n30| \\t// go doc -src fmt Println\\n31| \\n32| \\t// Println is just an exported function from\\n33| \\t// \\\"fmt\\\" package\\n34| \\n35| \\t// Exported = First Letter is uppercase\\n36| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n37| \\n38| \\t// Go cannot call Println function by itself.\\n39| \\t// It only calls `func main` automatically.\\n40| \\t// It only calls `func main` automatically.\\n41| \\n42| \\t// -----\\n43| \\n44| \\t// Go supports Unicode characters in string literals\\n45| \\t// And also in source-code: KÖSTEBEK!\\n46| \\t//\\n47| \\t// Because: Literal ~= Source Code\\n48| \\n49| \\t// EXERCISE: Remove the comments from below --> //\\n```\\n\\n## Result 3: github.com/Go-Learning-Land/Go-Learning-Archive/GolangBasics/learngo/translation/spanish/02-tu-primer-programa/main.go\\n\\nURL: /github.com/Go-Learning-Land/Go-Learning-Archive/-/blob/GolangBasics/learngo/translation/spanish/02-tu-primer-programa/main.go\\n\\n```\\n13| /*\\n14| Esto es un comentario multilinea.\\n15| \\n16| la palabra import hace que otro paquete este disponible\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" es especial.\\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Despues de compilar el codigo,\\n31| // Go ejecutara esta funcion primero.\\n32| func main() {\\n33| \\t// despues del: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Despues de compilar el codigo,\\n31| func main() {\\n32| func main() {\\n33| \\t// despues del: import \\\"fmt\\\"\\n34| \\t// La funcion Println del paquete \\\"fmt\\\" estara disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una funcion exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n```\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| // func main crea un punto de inicio para Go\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Despues de compilar el codigo,\\n31| // Go ejecutara esta funcion primero.\\n32| func main() {\\n33| \\t// despues del: import \\\"fmt\\\"\\n34| \\t// La funcion Println del paquete \\\"fmt\\\" estara disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n```\\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una funcion exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n42| \\t// Exportada = Primera letra en mayuscula\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go no puede llamar a la funcion Println por si mismo.\\n46| \\t// Solo llama a la `func main` de forma automatica.\\n47| \\t// Solo llama a la `func main` de forma automatica.\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go soporta caracteres unicode en cadenas de texto literal\\n52| \\t// y tambien en el codigo fuente: KÖSTEBEK!\\n53| \\t//\\n54| \\t// Porque: Literal ~= Codigo Fuente\\n55| }\\n56| \\n```\\n\\n## Result 4: github.com/inancgumus/learngo/translation/chinese/02-编写第一个程序/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/translation/chinese/02-%E7%BC%96%E5%86%99%E7%AC%AC%E4%B8%80%E4%B8%AA%E7%A8%8B%E5%BA%8F/main.go\\n\\n```\\n13| /*\\n14| 这是一个多行注释\\n15| \\n16| import 关键字引入另一个包\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" 是特殊的\\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| func main() {\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n```\\n\\n```\\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n42| \\t// Exported = 首字母大写\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go 本身不能调用 Println 函数\\n46| \\t// Go 只能自动地调用 `func main`\\n47| \\t// Go 只能自动地调用 `func main`\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go 字符串支持 Unicode 字符\\n52| \\t// 源代码同时也支持: KÖSTEBEK!\\n53| \\t//\\n54| \\t// 因为: 字面值 ~= 源代码\\n55| \\n56| \\t// 练习: 删除下行注释的 --> //\\n```\\n\\n```\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| // func main 对于 Go 而言就是开始执行的位置\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n## Result 5: github.com/inancgumus/learngo/first/first/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/first/first/main.go\\n\\n```\\n13| /*\\n14| This is a multi-line comment.\\n15| \\n16| import keyword makes another package available\\n17| for this .go \\\"file\\\".\\n18| \\n19| import \\\"fmt\\\" lets you access fmt package's functionality\\n20| here in this file.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" is special.\\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| // Go runtime will first run this function\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| func main() {\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n34| \\t// Println function of \\\"fmt\\\" package becomes available\\n35| \\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println is just an exported function from\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n```\\n17| for this .go \\\"file\\\".\\n18| \\n19| import \\\"fmt\\\" lets you access fmt package's functionality\\n20| here in this file.\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| // func main creates a starting point for Go\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| // Go runtime will first run this function\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n34| \\t// Println function of \\\"fmt\\\" package becomes available\\n35| \\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n```\\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println is just an exported function from\\n40| \\t// \\\"fmt\\\" package\\n41| \\n42| \\t// Exported = First Letter is uppercase\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go cannot call Println function by itself.\\n46| \\t// It only calls `func main` automatically.\\n47| \\t// It only calls `func main` automatically.\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go supports Unicode characters in string literals\\n52| \\t// And also in source-code: KÖSTEBEK!\\n53| \\n54| \\t// EXERCISE: Remove the comments from below --> //\\n55| \\t// fmt.Println(\\\"Merhaba Köstebek!\\\")\\n56| \\n```\\n\\n## Result 6: github.com/inancgumus/learngo/translation/spanish/02-tu-primer-programa/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/translation/spanish/02-tu-primer-programa/main.go\\n\\n```\\n13| /*\\n14| Esto es un comentario multilinea.\\n15| \\n16| la palabra import hace que otro paquete este disponible\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" es especial.\\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Después de compilar el codigo,\\n31| // Go ejecutará esta función primero.\\n32| func main() {\\n33| \\t// después del: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| //\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Después de compilar el codigo,\\n31| func main() {\\n32| func main() {\\n33| \\t// después del: import \\\"fmt\\\"\\n34| \\t// La función Println del paquete \\\"fmt\\\" estará disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una función exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n```\\n17| para este \\\"archivo\\\" .go.\\n18| \\n19| import \\\"fmt\\\" te permite acceder a la funcionaliddad del paquete fmt\\n20| en este archivo.\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" es especial.\\n25| //\\n26| // Go tiene que saber por donde empezar\\n27| // func main crea un punto de inicio para Go\\n28| // func main crea un punto de inicio para Go\\n29| //\\n30| // Después de compilar el codigo,\\n31| // Go ejecutará esta función primero.\\n32| func main() {\\n33| \\t// después del: import \\\"fmt\\\"\\n34| \\t// La función Println del paquete \\\"fmt\\\" estará disponible\\n35| \\n36| \\t// Lee sobre ella escribiendo los siguiente en la consola:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n```\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println es simplemente una función exportada de:\\n40| \\t// \\\"fmt\\\" package\\n41| \\n42| \\t// Para poder exportar una función tendrás que escribir el\\n43| \\t// primer caracter del nombre de la función en mayúscula.\\n44| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n45| \\n46| \\t// Go no puede llamar a la función Println por si mismo.\\n47| \\t// Solo llama a la `func main` de forma automatica.\\n48| \\t// Solo llama a la `func main` de forma automatica.\\n49| \\n50| \\t// -----\\n51| \\n52| \\t// Go soporta caracteres unicode en cadenas de texto literal\\n53| \\t// y tambien en el codigo fuente: KÖSTEBEK!\\n54| \\t//\\n55| \\t// Porque: Literal ~= Codigo Fuente\\n56| }\\n57| \\n```\\n\\n## Result 7: github.com/Go-Learning-Land/Go-Learning-Archive/GolangBasics/learngo/translation/chinese/02-编写第一个程序/main.go\\n\\nURL: /github.com/Go-Learning-Land/Go-Learning-Archive/-/blob/GolangBasics/learngo/translation/chinese/02-%E7%BC%96%E5%86%99%E7%AC%AC%E4%B8%80%E4%B8%AA%E7%A8%8B%E5%BA%8F/main.go\\n\\n```\\n13| /*\\n14| 这是一个多行注释\\n15| \\n16| import 关键字引入另一个包\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" 是特殊的\\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| //\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| func main() {\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n```\\n\\n```\\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println 是一个 exported 函数来自\\n40| \\t// \\\"fmt\\\" 包\\n41| \\n42| \\t// Exported = 首字母大写\\n43| \\tfmt.Println(\\\"Hello Gopher!\\\")\\n44| \\n45| \\t// Go 本身不能调用 Println 函数\\n46| \\t// Go 只能自动地调用 `func main`\\n47| \\t// Go 只能自动地调用 `func main`\\n48| \\n49| \\t// -----\\n50| \\n51| \\t// Go 字符串支持 Unicode 字符\\n52| \\t// 源代码同时也支持: KÖSTEBEK!\\n53| \\t//\\n54| \\t// 因为: 字面值 ~= 源代码\\n55| \\n56| \\t// 练习: 删除下行注释的 --> //\\n```\\n\\n```\\n17| 对于这个 .go ”文件“\\n18| \\n19| import \\\"fmt\\\" 允许你使用 fmt 包的各种函数\\n20| 在这个文件\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" 是特殊的\\n25| //\\n26| // Go 必须知道从哪里开始执行\\n27| // func main 对于 Go 而言就是开始执行的位置\\n28| // func main 对于 Go 而言就是开始执行的位置\\n29| //\\n30| // 代码编译后，\\n31| // Go runtime 第一个执行该函数\\n32| func main() {\\n33| \\t// import \\\"fmt\\\" 之后\\n34| \\t// \\\"fmt\\\" 包中的 Println 函数就可以使用\\n35| \\n36| \\t// 在控制台输入下面的命令可以查看它的源代码:\\n37| \\t// go doc -src fmt Println\\n```\\n\\n## Result 8: github.com/inancgumus/learngo/02-write-your-first-program/main.go\\n\\nURL: /github.com/inancgumus/learngo/-/blob/02-write-your-first-program/main.go\\n\\n```\\n13| /*\\n14| This is a multi-line comment.\\n15| \\n16| import keyword makes another package available\\n17| for this .go \\\"file\\\".\\n18| \\n19| import \\\"fmt\\\" lets you access fmt package's functionality\\n20| here in this file.\\n21| */\\n22| import \\\"fmt\\\"\\n23| // \\\"func main\\\" is special.\\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| // Go runtime will first run this function\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n```\\n\\n```\\n21| */\\n22| import \\\"fmt\\\"\\n23| \\n24| // \\\"func main\\\" is special.\\n25| //\\n26| // Go has to know where to start\\n27| //\\n28| // func main creates a starting point for Go\\n29| //\\n30| // After compiling the code,\\n31| func main() {\\n32| func main() {\\n33| \\t// after: import \\\"fmt\\\"\\n34| \\t// Println function of \\\"fmt\\\" package becomes available\\n35| \\n36| \\t// Look at what it looks like by typing in the console:\\n37| \\t// go doc -src fmt Println\\n38| \\n39| \\t// Println is just an exported function from\\n40| \\t// \\\"fmt\\\" package\\n41| \\n```\\n\\n\",\"tool_call_id\":\"call_usUSsyqHEqfNtcKQR2sSKbZn\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -194,239 +200,99 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"xynOe1Q3AQ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Found"},"finish_reason":null}],"usage":null,"obfuscation":"4pFS2t3"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" "},"finish_reason":null}],"usage":null,"obfuscation":"leudtyrxHOi"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"30"},"finish_reason":null}],"usage":null,"obfuscation":"38c8xSsqAE"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" matches"},"finish_reason":null}],"usage":null,"obfuscation":"Vjv7"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" ("},"finish_reason":null}],"usage":null,"obfuscation":"N5WYe0gXpS"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"limit"},"finish_reason":null}],"usage":null,"obfuscation":"FdnLzdn"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" reached"},"finish_reason":null}],"usage":null,"obfuscation":"exvI"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":").\n"},"finish_reason":null}],"usage":null,"obfuscation":"0OrbsjTE"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"github"},"finish_reason":null}],"usage":null,"obfuscation":"tB9ZRx"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"fat78KTy"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/w"},"finish_reason":null}],"usage":null,"obfuscation":"jsxXcdSzQV"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"x"},"finish_reason":null}],"usage":null,"obfuscation":"f6j3VxcD79v"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-che"},"finish_reason":null}],"usage":null,"obfuscation":"kh7LzyE6"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"val"},"finish_reason":null}],"usage":null,"obfuscation":"PfHXLhiX2"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ier"},"finish_reason":null}],"usage":null,"obfuscation":"2Kb3aYYdR"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"dPC5q04lIQf"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Go"},"finish_reason":null}],"usage":null,"obfuscation":"bVqJwg9xCz"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"DdFpqVkuJvv"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Notes"},"finish_reason":null}],"usage":null,"obfuscation":"xRdxdX3"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/-"},"finish_reason":null}],"usage":null,"obfuscation":"asz98PNuXu"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"vf5q1sQAq51"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"blob"},"finish_reason":null}],"usage":null,"obfuscation":"U1jHYbUG"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"vyXJ6Tjgrmu"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"99"},"finish_reason":null}],"usage":null,"obfuscation":"qcdy3SEE9c"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"~"},"finish_reason":null}],"usage":null,"obfuscation":"ZvBzTcdLOee"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"参考"},"finish_reason":null}],"usage":null,"obfuscation":"drpu26slsx"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"资料"},"finish_reason":null}],"usage":null,"obfuscation":"KpfOsHrUDc"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"bntBbW4sWhx"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"201"},"finish_reason":null}],"usage":null,"obfuscation":"duKvmbKB8"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"7"},"finish_reason":null}],"usage":null,"obfuscation":"dAe4ZWT2LS7"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"~"},"finish_reason":null}],"usage":null,"obfuscation":"XIL7T3q3ZFu"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"A"},"finish_reason":null}],"usage":null,"obfuscation":"ZJPiOnS1zXZ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"%"},"finish_reason":null}],"usage":null,"obfuscation":"B9MAGSgXe6z"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"20"},"finish_reason":null}],"usage":null,"obfuscation":"4LUY2LW9YJ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Huge"},"finish_reason":null}],"usage":null,"obfuscation":"Uu6tnnLW"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"%"},"finish_reason":null}],"usage":null,"obfuscation":"TTAUjI31caq"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"20"},"finish_reason":null}],"usage":null,"obfuscation":"Vb87lDBuce"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Number"},"finish_reason":null}],"usage":null,"obfuscation":"eGr1mv"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"%"},"finish_reason":null}],"usage":null,"obfuscation":"CAQlW8awPVz"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"20"},"finish_reason":null}],"usage":null,"obfuscation":"43GhuezTaK"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"of"},"finish_reason":null}],"usage":null,"obfuscation":"AXXjOi1WGX"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"%"},"finish_reason":null}],"usage":null,"obfuscation":"vkqxRlZnnLZ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"20"},"finish_reason":null}],"usage":null,"obfuscation":"cxImH7jbgP"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Go"},"finish_reason":null}],"usage":null,"obfuscation":"Z0218SsvjZ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"%"},"finish_reason":null}],"usage":null,"obfuscation":"SRNNZSUteJt"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"20"},"finish_reason":null}],"usage":null,"obfuscation":"4EXKMO0kFP"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Examples"},"finish_reason":null}],"usage":null,"obfuscation":"8LOS"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/c"},"finish_reason":null}],"usage":null,"obfuscation":"uEdASexoe8"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"odes"},"finish_reason":null}],"usage":null,"obfuscation":"LWnkBEfQ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"6cycQx5QVzA"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"first"},"finish_reason":null}],"usage":null,"obfuscation":"2mlRu7U"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"ychMqNe4a6V"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"first"},"finish_reason":null}],"usage":null,"obfuscation":"0O2Soxl"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"mCA0cR2"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"e8039dRpl"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"VJuHw3IFmk"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"github"},"finish_reason":null}],"usage":null,"obfuscation":"KRFErC"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"oaxJAuBd"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/in"},"finish_reason":null}],"usage":null,"obfuscation":"4YtmHPPCX"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"anc"},"finish_reason":null}],"usage":null,"obfuscation":"6gSbOTVpB"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"gum"},"finish_reason":null}],"usage":null,"obfuscation":"y0QTY1HuZ"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"us"},"finish_reason":null}],"usage":null,"obfuscation":"pOqleuDnhA"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"nuuJkoh5Ak7"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"lear"},"finish_reason":null}],"usage":null,"obfuscation":"FR5K1Zxi"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ngo"},"finish_reason":null}],"usage":null,"obfuscation":"WVEVpzNkP"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/-"},"finish_reason":null}],"usage":null,"obfuscation":"H4K2m31hHV"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"VfYRJqHmD6L"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"blob"},"finish_reason":null}],"usage":null,"obfuscation":"SatgHXS7"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"NoqGdndifEV"} - - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"02"},"finish_reason":null}],"usage":null,"obfuscation":"KuIfZF1HGL"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"5QodNI9GUx"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-write"},"finish_reason":null}],"usage":null,"obfuscation":"DNTDzg"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"30"},"finish_reason":null}],"usage":null,"obfuscation":"Lliviu8vMe"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-your"},"finish_reason":null}],"usage":null,"obfuscation":"9kXJYQK"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" matches"},"finish_reason":null}],"usage":null,"obfuscation":"AGJG"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-first"},"finish_reason":null}],"usage":null,"obfuscation":"R4kGTH"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"."},"finish_reason":null}],"usage":null,"obfuscation":"pxYOuK134LW"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-program"},"finish_reason":null}],"usage":null,"obfuscation":"TQz1"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" Examples"},"finish_reason":null}],"usage":null,"obfuscation":"91Q"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"95d5pzL"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":\n"},"finish_reason":null}],"usage":null,"obfuscation":"hos2YF3BM"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"ZYRYm5LhA"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"uRGlflxFAvF"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"BPeSvTSSch"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" github"},"finish_reason":null}],"usage":null,"obfuscation":"j8GOg"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"github"},"finish_reason":null}],"usage":null,"obfuscation":"6KebPH"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"XDyF9isA"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"D7V9IA4O"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/in"},"finish_reason":null}],"usage":null,"obfuscation":"6oVsvepRf"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"q0tHZlplewA"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"anc"},"finish_reason":null}],"usage":null,"obfuscation":"687kcHDNv"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Go"},"finish_reason":null}],"usage":null,"obfuscation":"LbalRfPwC1"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"gum"},"finish_reason":null}],"usage":null,"obfuscation":"5ZcfjlGVF"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-L"},"finish_reason":null}],"usage":null,"obfuscation":"Kw10ybWl8V"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"us"},"finish_reason":null}],"usage":null,"obfuscation":"JaRHTfCcbT"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"earning"},"finish_reason":null}],"usage":null,"obfuscation":"B0pOI"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"Okmfxk8jZld"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-L"},"finish_reason":null}],"usage":null,"obfuscation":"m79imykhqD"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"lear"},"finish_reason":null}],"usage":null,"obfuscation":"vq29A8xn"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"and"},"finish_reason":null}],"usage":null,"obfuscation":"55u9bpk0c"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ngo"},"finish_reason":null}],"usage":null,"obfuscation":"dal19cj4w"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"R57LS90F6aj"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"9mkCWkOZuer"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Go"},"finish_reason":null}],"usage":null,"obfuscation":"V0R8JHUeOq"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"02"},"finish_reason":null}],"usage":null,"obfuscation":"ZpBYSsCLqx"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-L"},"finish_reason":null}],"usage":null,"obfuscation":"YAmnJ2An4L"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-write"},"finish_reason":null}],"usage":null,"obfuscation":"BOfUfJ"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"earning"},"finish_reason":null}],"usage":null,"obfuscation":"360O6"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-your"},"finish_reason":null}],"usage":null,"obfuscation":"5iz3r89"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"bUJTScjxVue"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-first"},"finish_reason":null}],"usage":null,"obfuscation":"GSlvJA"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Archive"},"finish_reason":null}],"usage":null,"obfuscation":"kAnn9"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-program"},"finish_reason":null}],"usage":null,"obfuscation":"3pwu"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/-"},"finish_reason":null}],"usage":null,"obfuscation":"CfFNXlrfdJ"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"0vF8V6U"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"XszTvr26Grg"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"UoKWuxpbS"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"blob"},"finish_reason":null}],"usage":null,"obfuscation":"xM40mTKS"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"UmeZaH4IVTq"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/G"},"finish_reason":null}],"usage":null,"obfuscation":"2ttJyfSEI2"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"32"},"finish_reason":null}],"usage":null,"obfuscation":"IUiAhZ1kyM"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"olang"},"finish_reason":null}],"usage":null,"obfuscation":"WJtjSae"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"\n"},"finish_reason":null}],"usage":null,"obfuscation":"15qYlSZTEe"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Basics"},"finish_reason":null}],"usage":null,"obfuscation":"tv8md1"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"NDe3tlC9JJ6"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"86wYoEAvhjG"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":" github"},"finish_reason":null}],"usage":null,"obfuscation":"pP3Mv"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"lear"},"finish_reason":null}],"usage":null,"obfuscation":"YAihhcg8"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".com"},"finish_reason":null}],"usage":null,"obfuscation":"H8Ocde8w"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ngo"},"finish_reason":null}],"usage":null,"obfuscation":"FKOrj8XXH"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/in"},"finish_reason":null}],"usage":null,"obfuscation":"4O0pSEA2z"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"TN6samlG4lq"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"anc"},"finish_reason":null}],"usage":null,"obfuscation":"x0uKLtsqb"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"translation"},"finish_reason":null}],"usage":null,"obfuscation":"t"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"gum"},"finish_reason":null}],"usage":null,"obfuscation":"3ZKoJhRyH"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/ch"},"finish_reason":null}],"usage":null,"obfuscation":"mTGHwy5F7"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"us"},"finish_reason":null}],"usage":null,"obfuscation":"hsPmEMV2pp"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"inese"},"finish_reason":null}],"usage":null,"obfuscation":"DHZQYZH"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"6qKx6eo6s5M"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"4ftvG7OwgEr"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"lear"},"finish_reason":null}],"usage":null,"obfuscation":"7TjYybUV"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"02"},"finish_reason":null}],"usage":null,"obfuscation":"Qib5O7cXpy"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"ngo"},"finish_reason":null}],"usage":null,"obfuscation":"cGVQVX7lJ"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"-"},"finish_reason":null}],"usage":null,"obfuscation":"CVBW0XCWoZT"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"Z6IwVJQOCvk"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"编"},"finish_reason":null}],"usage":null,"obfuscation":"rzkLNuUmY4k"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"first"},"finish_reason":null}],"usage":null,"obfuscation":"sk8f6oP"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"写"},"finish_reason":null}],"usage":null,"obfuscation":"osoNMO3tvsy"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/"},"finish_reason":null}],"usage":null,"obfuscation":"z6hr7pEE9Ju"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"第"},"finish_reason":null}],"usage":null,"obfuscation":"YM4Sc5mRSDc"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"first"},"finish_reason":null}],"usage":null,"obfuscation":"x9RY2em"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"一个"},"finish_reason":null}],"usage":null,"obfuscation":"Ar0QPjM2BG"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"gZpNTG9"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"程序"},"finish_reason":null}],"usage":null,"obfuscation":"hylnFmrGAF"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"PrNiu0zKC"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"/main"},"finish_reason":null}],"usage":null,"obfuscation":"JdV4zYd"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":":"},"finish_reason":null}],"usage":null,"obfuscation":"fFZtRkcTMx8"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":".go"},"finish_reason":null}],"usage":null,"obfuscation":"PFYzaIlKI"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"32"},"finish_reason":null}],"usage":null,"obfuscation":"HZRON4rYHm"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"XZVvkY"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"RxQxtf"} - data: {"id":"chatcmpl-CVfO6rAmCfSkNhKGrySj4DnHJdCK0","object":"chat.completion.chunk","created":1761663298,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":16163,"completion_tokens":1019,"total_tokens":17182,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":896,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"2lc1TgjNZ92lWL"} + data: {"id":"chatcmpl-CVgwvS3Ne5F2MhEOz4or65AvQrTXY","object":"chat.completion.chunk","created":1761669301,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":15917,"completion_tokens":629,"total_tokens":16546,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":576,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"VFVdUxWDiGkdnfY"} data: [DONE] @@ -435,4 +301,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 13.845728042s + duration: 10.3609635s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/update_a_file.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/update_a_file.yaml index 2dd7ae02ba3b628ce8a8dea0ef8994a18dadffb8..f1077fb6c1e938908cda9e1a9ec7e09a9a9d628d 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/update_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/update_a_file.yaml @@ -24,25 +24,31 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"j63YGLvoDyULXC"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"jj1v9cE4mDQW80"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Modify"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"K8DT17VSSx"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Modify"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"DiWZ4H7kp4"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" print"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"tXmZkVBFZP"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" main"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"gFWhcNpwqwN"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" message"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"f0VQqrrz"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":".go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"sgbxkW9VDE3wA"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" in"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"wsiSpiua2O2eU"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" to"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"nJ1xnorC4W9Cx"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" main"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"7M2nevlS68R"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Print"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"DwUaVp9TOq"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":".go"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"9S8yEPgw6A7zp"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" \""},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"lM0l9C7dyubp8"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" file"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"J0QIBlhlREn"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Hello"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"NM93PoPdNu3"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"yaRllmB9Pp"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" from"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"yb22EDG36Hz"} - data: {"id":"chatcmpl-CVfMOs0RaYbcYeRux2a0nPOEcnU8N","object":"chat.completion.chunk","created":1761663192,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":139,"completion_tokens":7,"total_tokens":146,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"PX4Qck3VG26qVg"} + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Crush"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"ylLR3m2r0J"} + + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"\""},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"LQaxTaSVULPSTI"} + + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"ubOo9RIFcZ"} + + data: {"id":"chatcmpl-CVguFCA8w4DxmnGYZVsRMg3YrPoI5","object":"chat.completion.chunk","created":1761669135,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":139,"completion_tokens":10,"total_tokens":149,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"kVBCQGb62GtJ9"} data: [DONE] @@ -51,15 +57,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 626.675125ms + duration: 1.754230125s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44167 + content_length: 42950 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -75,65 +81,65 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_pqnbjrnfunsPMfQnAZLJrmfp","type":"function","function":{"name":"ls","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"pfs6ZpI5Hn8sQk"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_tS9FCGm1COVRXkRce95uvc1V","type":"function","function":{"name":"ls","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"FgpQYRy8QOJgF3"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"i2zXIDmlvzFvbvi"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"ASYNwG2GC9oViyQ"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5yWwZIQ2h61q3L"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dVEEJOVd7FQPNs"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IzQVKDYJs5DsCLD"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2BxBye2NSmvj4HS"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YEOewx9snani4F"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rGjM6VLU5Igg6k"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2D2PYZ6pfjxIDEf"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r5hrAE78dxgs4SJ"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CDd5pddfgV7jfBT"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UBnnKsyNfvuBrus"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bU6DWXmTRC1fbFX"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cE78UBDvSxUc3sA"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"4hPUihmYxOYYQ"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uBHAPYfKgPv6o"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dZSGmbbUIvXc1"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ARId03SkPMGsD"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qFTOg8u0Dj7gK"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zITwrwtyiFZZl"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lmmjkMNRfKc4q"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vL55LKI4lKGEf"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Qnv4h8a8th9VG"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"f9kSBYmi87wPA"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"h"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"G"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vl1l92Ss639"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pp3QX3sHi01"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IF2nrW9557620"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rZMlE1hyivmOp"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"D5W4nnkvqb9lJB4"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"EVpk8ummLiIjLaZ"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"YFqty5WspbBfzKL"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \""}}]},"finish_reason":null}],"usage":null,"obfuscation":"AR0MDF9GCBuSVR4"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"depth"}}]},"finish_reason":null}],"usage":null,"obfuscation":"W46729HhLdVeF"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"depth"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NyqDrM8Q046z1"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NxylFUqg1btMXmk"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZIgHcChZ6TIXLBH"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"5"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" "}}]},"finish_reason":null}],"usage":null,"obfuscation":"h"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"2"}}]},"finish_reason":null}],"usage":null,"obfuscation":"l"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"2"}}]},"finish_reason":null}],"usage":null,"obfuscation":"u"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"R"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"E"} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMP0I9C5sOUozIXjI6RCUEmPJ4S","object":"chat.completion.chunk","created":1761663193,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9744,"completion_tokens":109,"total_tokens":9853,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":64,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"QT"} + data: {"id":"chatcmpl-CVguF2QvEv4ovxLLsBxOwPwyfXvE3","object":"chat.completion.chunk","created":1761669135,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9493,"completion_tokens":173,"total_tokens":9666,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":128,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"d"} data: [DONE] @@ -142,15 +148,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 2.521306958s + duration: 4.730460042s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44552 + content_length: 43335 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -166,59 +172,59 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_UOJzgz3klg9lXgBH347jYRvM","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"SKPJdt9OHVNt"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_CqMXE6SDDWqHO12VSsyQI7Tj","type":"function","function":{"name":"view","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"NMAt0KMQZa0J"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"mp5S5IvcRk3GLix"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"pyBezPLuuz4PmcQ"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ARxKxyyYNnd24P"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GHhMmQwbFRJcjK"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Mfamr6P07Y8om"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"R4sXpcOvhHcl4"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3JKBIAtiOoPPZCp"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A9zfJTarfQpZt0h"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kPc99JHfgF0yIQf"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Vbm2XJCwI4pABeP"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"mggHTTAmLZ4Go5e"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"j5XxqCNmPHv1y2i"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Af6JeXpd1PoX3bW"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BQVWiAsZW1pfAqE"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GS4h6kEpQDjhsaE"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"H60QsuUk6I3Tfsh"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nXlciV7XtDbTs"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dwxYilHDZ48w6"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"SBWNTVVu5CV5i"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"aSIzMlhsc58Lz"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CuKRwlC8AIhXc"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"4qolm1Lj0fqoM"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zNmWh3VOm346B"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"x5iFZrCywDy62"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3BDrA6xABEts5"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ftmZdXzmSN3ad"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"W"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tBrupDyw7qv"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Wk2T0AcweMW"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"j7RWaapebstO6"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UeuQYqncA6Blx"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Binn6pGcECjBm"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rDsfEYeiOFcpN"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"o81LRcD1KEMJu6V"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"TTwLZXWxsSnbOr9"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D7DFv2pfcenjQtQ"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yTwjimXchS8AC8y"} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMRNazf7gHtmqcxeMwI3WTmhOnx","object":"chat.completion.chunk","created":1761663195,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9821,"completion_tokens":35,"total_tokens":9856,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"ttxj"} + data: {"id":"chatcmpl-CVguMd2Ag0CXaquskErP6r9F2J1vv","object":"chat.completion.chunk","created":1761669142,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9570,"completion_tokens":35,"total_tokens":9605,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"lc57"} data: [DONE] @@ -227,15 +233,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.374454834s + duration: 3.216867833s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45000 + content_length: 43783 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_UOJzgz3klg9lXgBH347jYRvM\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_UOJzgz3klg9lXgBH347jYRvM\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_CqMXE6SDDWqHO12VSsyQI7Tj\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_CqMXE6SDDWqHO12VSsyQI7Tj\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -251,185 +257,151 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_q4UClQoKojXGsPTWSdWUjmgA","type":"function","function":{"name":"edit","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"eqw21yCxb7bd"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"CWpA1icisXA3qbF"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"RRYO4SoG8tNAD2"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HgzVgBINYahLo"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"YirXb5yQ8KceeAq"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XHtYsCQRsqV9QBe"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BnIOWZGBuyWnRDZ"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lxZhY9jCU4BcCQs"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hopucoQNo9xKd8A"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_lFmlMx6rc90jXuWLpq25i5bd","type":"function","function":{"name":"edit","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"yioveHHaPWqw"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GYGgZs8TUbYoQ"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"JrEOuMTvsEO8Ho8"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MG8LK2mn0ClPJ"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7nejx9ibMuX960"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"eLCd8peAZrXaO"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sH79En0xlU2gw"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"He4kP4CzvDzUX"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BWWVCNQlxNMMa7z"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Hrn9SLLHeWvEo"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QU1gb2JUt8Nf6dX"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cTsqlzyOUejunEK"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yvPJbEvHJpyfWDE"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OvVw9P2tCRLcYkL"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7dvusf3o2BlQX"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D7lddvASlq9bQ"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7WjaMNcuhWm"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0GDDFhDMzjH7N"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XRvCJZlLcM46U"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zY3BLb0Zhc9IM"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fJpdrkGEpvksG"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GJ8ZCXISv0wwO"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QTbjNOkSs7psoZG"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"hz4mGaDr6dLJr"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old"}}]},"finish_reason":null}],"usage":null,"obfuscation":"m3JThHCk1SKvZ7U"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"F8KyYWD5JgQ"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"x"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"63NQ413jQOulf"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zX6HryTIZbQ"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"package"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MimSaqUdbzq"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"SJ6MJbw8wAfg3"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"JpWTVQ9wPGu2v"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Vx0ShpcapEMioa0"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kFmULmhlfxbxr"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"QYbElzVn4SeKdss"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uzGjssSVM0D856p"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"import"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GspJIyUuJ0Ua"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"49xQgnl70CvS2"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"H9ti8bZw99MOp"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"old"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hv8XJWedPvC9QGY"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2kSmiU10AAXhYTz"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xf3w2Vxbhkr"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"loCBa5nYHUqV"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Pxd5njL3vPjz4"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"w"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xLC9RXi17vhECr"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gtn3bpX8K0U6ixs"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"HOfPog1jcU1Mt"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"i8hRfj0P9aQN68"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9OxbDIv8FkFA5"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yKkkgmTfayD9gt"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"D"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lS3R6hBO5RWjTE"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sGHJ6ys5fzkbEHR"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"w"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"VIPktuVZQjQO19D"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"mBFSyRZu8eIoXK0"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5ZVjOokVR8"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kqLIFxjzsBuBkC2"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"oYJY0mCTT6pGV"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pGxljvwGAH"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"18fwKcSQx8sGx"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"JaN6aGqycXIKC"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"4"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"obXYmLfcywX8z"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" World"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5ReQs0j0Hjzv"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":","}}]},"finish_reason":null}],"usage":null,"obfuscation":"i"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wPTGz2StW3qlah0"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" World"}}]},"finish_reason":null}],"usage":null,"obfuscation":"p3ma3YkuxaVq"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oxqSR3Km06HQ7Yu"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"!\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0o7Q9dK4Euc39ui"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xaJpHNW3etkwdAN"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"l"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"n"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"PfCIFYwZHwkMO"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CEmSPzq0A5HKZDb"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"W4nF29DXKxpuHgt"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lNxL0ZNBv8q"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"rhwsZCTFTFysG"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"hEHSKXgUiTT6b"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Ou4qeS0zUZBKmeV"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5fHYxLyQmmat5x"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_string"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yQYLQ7ILfej"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"So7v2dqawDKxC"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Ia3OhHPOCTx4C"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"package"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hIXClBh6MkW"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"OUBrlB44H4WMcm"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"9eI1boCR5ePpW"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"r7lnPVk9TA3uffh"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qSj8pW5RyzB1sw8"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iIUSL6oaWm7Xqdu"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"atHZ5STEiPWkICw"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"import"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3szICLXtjnuc"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1dYGwwBAs2"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"4vSxDc1aNnjki"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"lzPNw3BDNZVud"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pcKsafVKeuDyXEo"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"NSk7Iz1rjQylD"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vyxBy2FkvKGa"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" from"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T8pnSChyCXT56"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"K"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7bpxawqXzCAL"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BCpKUWuBjqxoT7l"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tefnIyMcbwuOj"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"func"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qN3jwipbsjmti0"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uKa9d6BYZLUG3"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"M"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"()"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"H"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" {\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3n2SnaNP7Pkb7N"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"VWSEBkKIhrI9w"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"p"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"replace"}}]},"finish_reason":null}],"usage":null,"obfuscation":"U7WueUAJwP5"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\t"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sa1gkNhbBpQW4Ui"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_all"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LpSZnRs1VSOghS"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"fmt"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vNyez6F9nApdjDX"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zjdU817lhBKZimu"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".Println"}}]},"finish_reason":null}],"usage":null,"obfuscation":"UcAH1p2A0F"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"false"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iFQ8hcVALvs2o"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"(\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"DWFuNsAIAGEes"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"g"} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"hello"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bTLJIRZVgT20F"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" from"}}]},"finish_reason":null}],"usage":null,"obfuscation":"nFbKi7dCM1VG8"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" crush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"64kgSdd7OiHT"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\")"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z5cSCH2qpUQcX"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}\\"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pSgNRqDcgV3qZid"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"n"}}]},"finish_reason":null}],"usage":null,"obfuscation":"I"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pwpxm5OTbgKvyIh"} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - - data: {"id":"chatcmpl-CVfMTqWZMMy5vuJrMPLMMglgN0Ni3","object":"chat.completion.chunk","created":1761663197,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9918,"completion_tokens":98,"total_tokens":10016,"prompt_tokens_details":{"cached_tokens":9728,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"5nk"} + data: {"id":"chatcmpl-CVguQs9fy6TPCiEvc6dtbm8uvp8NQ","object":"chat.completion.chunk","created":1761669146,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9667,"completion_tokens":471,"total_tokens":10138,"prompt_tokens_details":{"cached_tokens":9600,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":384,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} data: [DONE] @@ -438,15 +410,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.49178325s + duration: 8.481338125s - id: 4 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45657 + content_length: 44372 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_UOJzgz3klg9lXgBH347jYRvM\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_UOJzgz3klg9lXgBH347jYRvM\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_q4UClQoKojXGsPTWSdWUjmgA\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\",\\\"old_string\\\":\\\"package main\\\\n\\\\nimport \\\\\\\"fmt\\\\\\\"\\\\n\\\\nfunc main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\\n\\\",\\\"new_string\\\":\\\"package main\\\\n\\\\nimport \\\\\\\"fmt\\\\\\\"\\\\n\\\\nfunc main() {\\\\n\\\\tfmt.Println(\\\\\\\"hello from crush\\\\\\\")\\\\n}\\\\n\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_q4UClQoKojXGsPTWSdWUjmgA\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_CqMXE6SDDWqHO12VSsyQI7Tj\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_CqMXE6SDDWqHO12VSsyQI7Tj\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_lFmlMx6rc90jXuWLpq25i5bd\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\",\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"hello from crush\\\\\\\")\\\\n}\\\",\\\"replace_all\\\":false}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_lFmlMx6rc90jXuWLpq25i5bd\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -462,85 +434,83 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_ZMyX9QW5TvFoiJsah7IAecSX","type":"function","function":{"name":"bash","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"sDHSyIceXtmq"} - - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"F1q7wtEv4HwICza"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_XerJaZ4DUKNiG6kpLChHgYAJ","type":"function","function":{"name":"bash","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"mVinnF2TPBC1"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"command"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pv8L7SmYUZw"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"XDYkuPnvuwRDgjG"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"TLYAp6ZFjd5kl"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"command"}}]},"finish_reason":null}],"usage":null,"obfuscation":"uryEwBLRMEF"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"h3O5HPzL4HFXb"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" run"}}]},"finish_reason":null}],"usage":null,"obfuscation":"K2YhClJ72TQhO2"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"go"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" /"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" run"}}]},"finish_reason":null}],"usage":null,"obfuscation":"jbW7oJ37cOWysK"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FFK7ZYqoNRUNUkh"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" /"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5rRWKqM7SrA3E9B"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lTHxOCjXNfAGbLf"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qsHrqCm8QSHSYjw"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"raAF9VA9JBMjAiJ"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ZqWsEKFlbkOrt"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"vBUWWRzTmm4Mogh"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"GcaVPVnxdNTqL"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"b8Mcq4GGzVAMn"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"W5oZMrdiDzHo1"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LYzXtJhhHAi8S"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WqdzLbyQilQB6"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"qK4bDDUQanhlK"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8gMr334lsV0zs"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"aYRiYZSb5EICl"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"i2wXsv95kkMrJ"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"f"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"e"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"S"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7rdBadeBBDq"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"5"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/update"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7TsLnhgCtVN"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FCtzCnpbwezSU"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_a"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"iG2mw1QAJXJFI"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8XOS9g02fvjsj"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"57ZcHelytJo8Ipg"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/main"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Bc2axvVqkrATi"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"4HMFknp5a6QLm"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hAqUfwo4FLGzbI2"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"description"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CGCptik"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"gVixc4iNd7Zks"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"U1pzsP3OHXYge"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"description"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Y3l5gBA"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Run"}}]},"finish_reason":null}],"usage":null,"obfuscation":"bNqVE9zm0OsqvOc"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"8CyaP06fjEMBk"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" the"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tXIZAcJ1Vzb5yL"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Run"}}]},"finish_reason":null}],"usage":null,"obfuscation":"tWeqXrwvElZw1Ut"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MR0zSyxFawlVcBX"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" the"}}]},"finish_reason":null}],"usage":null,"obfuscation":"fUOSzHOYTusx5u"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" program"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rs6sTWB1MR"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" Go"}}]},"finish_reason":null}],"usage":null,"obfuscation":"FzX31RfElWguiM4"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" to"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xnUi8iUtxIG5bzq"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" program"}}]},"finish_reason":null}],"usage":null,"obfuscation":"cqY5qy3sYI"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" verify"}}]},"finish_reason":null}],"usage":null,"obfuscation":"MLbiAA7iSOY"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" to"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sZqJ6DRPDvNPAbX"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" the"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z6xnEYBuNEezDR"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"V06DRzZ3jl5b4"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" new"}}]},"finish_reason":null}],"usage":null,"obfuscation":"IUiH6sOGd9buwZ"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" the"}}]},"finish_reason":null}],"usage":null,"obfuscation":"BJGc8daKxPpUGb"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" output"}}]},"finish_reason":null}],"usage":null,"obfuscation":"6iAfFC8yCmx"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" change"}}]},"finish_reason":null}],"usage":null,"obfuscation":"DNHKQI3yPFe"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ko3DWrQC1CWPIvK"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"ycPANB9C98DBCVF"} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfMULoitFvOeFsyGeFFzZHOGLVXD","object":"chat.completion.chunk","created":1761663198,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":10057,"completion_tokens":49,"total_tokens":10106,"prompt_tokens_details":{"cached_tokens":9984,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"It"} + data: {"id":"chatcmpl-CVguZApOSlZylxpK7ko7tYpyXU38c","object":"chat.completion.chunk","created":1761669155,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9789,"completion_tokens":48,"total_tokens":9837,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"uSWS"} data: [DONE] @@ -549,15 +519,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.206938083s + duration: 2.863020875s - id: 5 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 46112 + content_length: 44821 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_pqnbjrnfunsPMfQnAZLJrmfp\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_UOJzgz3klg9lXgBH347jYRvM\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_UOJzgz3klg9lXgBH347jYRvM\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_q4UClQoKojXGsPTWSdWUjmgA\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\",\\\"old_string\\\":\\\"package main\\\\n\\\\nimport \\\\\\\"fmt\\\\\\\"\\\\n\\\\nfunc main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\\n\\\",\\\"new_string\\\":\\\"package main\\\\n\\\\nimport \\\\\\\"fmt\\\\\\\"\\\\n\\\\nfunc main() {\\\\n\\\\tfmt.Println(\\\\\\\"hello from crush\\\\\\\")\\\\n}\\\\n\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_q4UClQoKojXGsPTWSdWUjmgA\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_ZMyX9QW5TvFoiJsah7IAecSX\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"go run /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\",\\\"description\\\":\\\"Run the Go program to verify the new output\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"hello from crush\\n\\n\\n/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\",\"tool_call_id\":\"call_ZMyX9QW5TvFoiJsah7IAecSX\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\\\", \\\"depth\\\": 2}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_tS9FCGm1COVRXkRce95uvc1V\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_CqMXE6SDDWqHO12VSsyQI7Tj\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_CqMXE6SDDWqHO12VSsyQI7Tj\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_lFmlMx6rc90jXuWLpq25i5bd\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\",\\\"old_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\",\\\"new_string\\\":\\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"hello from crush\\\\\\\")\\\\n}\\\",\\\"replace_all\\\":false}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_lFmlMx6rc90jXuWLpq25i5bd\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"call_XerJaZ4DUKNiG6kpLChHgYAJ\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"go run /tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file/main.go\\\",\\\"description\\\":\\\"Run the Go program to test the change\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"hello from crush\\n\\n\\n/tmp/crush-test/TestCoderAgent/openai-gpt-5/update_a_file\",\"tool_call_id\":\"call_XerJaZ4DUKNiG6kpLChHgYAJ\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -573,13 +543,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfMW5zZuvuTMfYZ0sY30P6pp97HE","object":"chat.completion.chunk","created":1761663200,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"htFcjICQTX"} + data: {"id":"chatcmpl-CVgub3kmmBXdDQjsOCrJvDKGMaS3U","object":"chat.completion.chunk","created":1761669157,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"7xS0bGcr8Z"} - data: {"id":"chatcmpl-CVfMW5zZuvuTMfYZ0sY30P6pp97HE","object":"chat.completion.chunk","created":1761663200,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"j2stbJkP"} + data: {"id":"chatcmpl-CVgub3kmmBXdDQjsOCrJvDKGMaS3U","object":"chat.completion.chunk","created":1761669157,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"4ys6xZhv"} - data: {"id":"chatcmpl-CVfMW5zZuvuTMfYZ0sY30P6pp97HE","object":"chat.completion.chunk","created":1761663200,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"lcvQR1"} + data: {"id":"chatcmpl-CVgub3kmmBXdDQjsOCrJvDKGMaS3U","object":"chat.completion.chunk","created":1761669157,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"Z8V0qT"} - data: {"id":"chatcmpl-CVfMW5zZuvuTMfYZ0sY30P6pp97HE","object":"chat.completion.chunk","created":1761663200,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":10142,"completion_tokens":4,"total_tokens":10146,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"9lJ"} + data: {"id":"chatcmpl-CVgub3kmmBXdDQjsOCrJvDKGMaS3U","object":"chat.completion.chunk","created":1761669157,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9873,"completion_tokens":4,"total_tokens":9877,"prompt_tokens_details":{"cached_tokens":9728,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"XZIDm"} data: [DONE] @@ -588,4 +558,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.226526s + duration: 6.551902917s diff --git a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/write_tool.yaml b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/write_tool.yaml index 66888afc4a609a08e15d725fe23cc0e747effb9d..2c8586dcbe490994bf13cc35764916285deeb2e1 100644 --- a/internal/agent/testdata/TestCoderAgent/openai-gpt-5/write_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openai-gpt-5/write_tool.yaml @@ -24,27 +24,23 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"4f43QAQEpCyyhA"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"5EvphZp8okqYWd"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Create"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"WzPkXnnYWT"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":"Create"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"eLBT4IOurj"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" a"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"FUNIZkXjgZUjqT"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" config"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"CqFbqukWz"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" config"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"eYwqpKVd7"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":".json"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"myte8RQlFNz"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":".json"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"pyq2dFHIsWL"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"dkMH83s4L5J"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" file"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"YKuyHiU8dAz"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" JSON"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"3ZuVhGd1ply"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" with"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"OUGm8fLbPJH"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" Content"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"7CBQfsP5"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" specific"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"VgDab2M"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"YNur7T841w"} - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{"content":" content"},"logprobs":null,"finish_reason":null}],"usage":null,"obfuscation":"MCtFcaVk"} - - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[{"index":0,"delta":{},"logprobs":null,"finish_reason":"stop"}],"usage":null,"obfuscation":"xYausAYJmP"} - - data: {"id":"chatcmpl-CVfOKiCaiMhxNnYtu0vIRZqI78RJW","object":"chat.completion.chunk","created":1761663312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":153,"completion_tokens":8,"total_tokens":161,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"FloMcS98ZyvPVc"} + data: {"id":"chatcmpl-CVgx6g7aleE2i4uANRjPfN3mw21Cb","object":"chat.completion.chunk","created":1761669312,"model":"gpt-4o-2024-08-06","service_tier":"default","system_fingerprint":"fp_65564d8ba5","choices":[],"usage":{"prompt_tokens":153,"completion_tokens":6,"total_tokens":159,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"P5Q1KQewc8yFLh"} data: [DONE] @@ -53,15 +49,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 652.074875ms + duration: 630.930208ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44204 + content_length: 42987 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -77,97 +73,97 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_EeXNb1uQZumswRoqQHgAgGAi","type":"function","function":{"name":"write","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"GgwahHrTwuj"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"call_EzgPZm2hPJwmVKxf82Eblmly","type":"function","function":{"name":"write","arguments":""}}],"refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"3QHV4e0M2vA"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"phxadoq15AWgSTn"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"GUZCrNJLVSFdf9V"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zvWAQrTBSQDAl2"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"file"}}]},"finish_reason":null}],"usage":null,"obfuscation":"sgwpQ6VStCkRIP"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"owj6kLggq3i4k"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_path"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lRYEsqCZWgmT5"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zWv2o1dRNH1SALc"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"gNLguMEheK0XzhB"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Bqi4p3t6FYjhLBp"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"/"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AJSFcYai7gBaHS6"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"KiMCZoGtScmQ5Qi"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"tmp"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zkS19XKIbhiiJE3"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"lQhxcPoPHO7ekMC"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/cr"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8SBdcszcqxEQTxc"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d7pnDcH909Klhue"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ush"}}]},"finish_reason":null}],"usage":null,"obfuscation":"yTTklQsMmEm8b3E"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"0tYzkQPkcL1u8"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"wZSvWbMeuqltC"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"EciDFyC1UcJWb"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/Test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"COykhqxFPoumu"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"CVmVku0SPP7Wo"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Coder"}}]},"finish_reason":null}],"usage":null,"obfuscation":"rFrSAd3QFeNV2"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"xUouo2u7VCNVh"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"Agent"}}]},"finish_reason":null}],"usage":null,"obfuscation":"1XWW8FgFzdBEc"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"S1TcvXZKWv4Zo"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/open"}}]},"finish_reason":null}],"usage":null,"obfuscation":"pTYmZu7pxReTd"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"ai"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-g"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"pt"}}]},"finish_reason":null}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"s"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"-"}}]},"finish_reason":null}],"usage":null,"obfuscation":"A"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"5"}}]},"finish_reason":null}],"usage":null,"obfuscation":"i"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/write"}}]},"finish_reason":null}],"usage":null,"obfuscation":"zofVilYV1ywb"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/write"}}]},"finish_reason":null}],"usage":null,"obfuscation":"l5wB1cIIMf3X"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"LtrPvVMriAiOA"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"_tool"}}]},"finish_reason":null}],"usage":null,"obfuscation":"J4N1VVdWWjWoj"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/config"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2lfaPE9g3ZM"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"/config"}}]},"finish_reason":null}],"usage":null,"obfuscation":"oNzLqFaDeCe"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".json"}}]},"finish_reason":null}],"usage":null,"obfuscation":"3WR7OkwO0wI8S"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":".json"}}]},"finish_reason":null}],"usage":null,"obfuscation":"hjUhYS8vfp5m0"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Z98Ca1NFNm6zQ"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"ImPJHKV7YPfYV"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"content"}}]},"finish_reason":null}],"usage":null,"obfuscation":"kZ0ALfIGPOr"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"content"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AoCaqcMOwNj"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"hOXLo8L9wEK16"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"nnIRlbfGILK0Y"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"RX8kCm3Snazso"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"axpOYD6QMJRCA"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"name"}}]},"finish_reason":null}],"usage":null,"obfuscation":"WUvuMzS9miMfwJ"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"name"}}]},"finish_reason":null}],"usage":null,"obfuscation":"myBnaXhjdbkZHn"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8csHp6m7wBImf"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"AYaHGhRcQU8PF"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"fBYfWkwPA3Cin"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"52WKUdm5cIzNg"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"dBxGEjCm2M7VZn"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"test"}}]},"finish_reason":null}],"usage":null,"obfuscation":"7o7EHonc4GibaQ"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"WlkXeUR9iozKn"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\","}}]},"finish_reason":null}],"usage":null,"obfuscation":"AmrEmAzWZXhLG"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"IIm11ntKccvJC"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"Ci0fkqgfcdl1N"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"version"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XFhnoM6Idss"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"version"}}]},"finish_reason":null}],"usage":null,"obfuscation":"X9wxEf9gzlY"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"03fJrIvL4B1Ja"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\":"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Xy3Z8gh1Woit1"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"XHPrUNzJ7UlUo"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":" \\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"hvNn34ytFEEMt"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1"}}]},"finish_reason":null}],"usage":null,"obfuscation":"2"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"1"}}]},"finish_reason":null}],"usage":null,"obfuscation":"q"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"."}}]},"finish_reason":null}],"usage":null,"obfuscation":"I"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"."}}]},"finish_reason":null}],"usage":null,"obfuscation":"F"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"0"}}]},"finish_reason":null}],"usage":null,"obfuscation":"T"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"0"}}]},"finish_reason":null}],"usage":null,"obfuscation":"8"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"."}}]},"finish_reason":null}],"usage":null,"obfuscation":"c"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"."}}]},"finish_reason":null}],"usage":null,"obfuscation":"m"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"0"}}]},"finish_reason":null}],"usage":null,"obfuscation":"z"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"0"}}]},"finish_reason":null}],"usage":null,"obfuscation":"Q"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"sic4AZ3QAS5nOo"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\\\""}}]},"finish_reason":null}],"usage":null,"obfuscation":"3cyi9fIIrlCfsY"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"c"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"d"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"XlPB0bNQVqJCicU"} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}],"usage":null,"obfuscation":"KJAMtYsPYZYrQPn"} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}],"usage":null,"obfuscation":""} - data: {"id":"chatcmpl-CVfOLmTWld5fph4GDWkryWFGC9IiN","object":"chat.completion.chunk","created":1761663313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9758,"completion_tokens":380,"total_tokens":10138,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":320,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":""} + data: {"id":"chatcmpl-CVgx7EZS0ftHXeVKpsuzhVOU85xQV","object":"chat.completion.chunk","created":1761669313,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9507,"completion_tokens":380,"total_tokens":9887,"prompt_tokens_details":{"cached_tokens":2688,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":320,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"Z"} data: [DONE] @@ -176,15 +172,15 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 10.190149041s + duration: 6.626818958s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44693 + content_length: 43476 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_EeXNb1uQZumswRoqQHgAgGAi\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool/config.json\\\",\\\"content\\\":\\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\"}\",\"name\":\"write\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nFile successfully written: /tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool/config.json\\n\",\"tool_call_id\":\"call_EeXNb1uQZumswRoqQHgAgGAi\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"call_EzgPZm2hPJwmVKxf82Eblmly\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool/config.json\\\",\\\"content\\\":\\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\"}\",\"name\":\"write\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nFile successfully written: /tmp/crush-test/TestCoderAgent/openai-gpt-5/write_tool/config.json\\n\",\"tool_call_id\":\"call_EzgPZm2hPJwmVKxf82Eblmly\",\"role\":\"tool\"}],\"model\":\"gpt-5\",\"max_completion_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -200,13 +196,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"chatcmpl-CVfOVarszi9WEHgzSz5mFugUiAnQ8","object":"chat.completion.chunk","created":1761663323,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"gxLFJzQdpU"} + data: {"id":"chatcmpl-CVgxGuhYP1UDMB0z0FHs1FIOt3DmR","object":"chat.completion.chunk","created":1761669322,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"role":"assistant","content":"","refusal":null},"finish_reason":null}],"usage":null,"obfuscation":"qr60f921Lt"} - data: {"id":"chatcmpl-CVfOVarszi9WEHgzSz5mFugUiAnQ8","object":"chat.completion.chunk","created":1761663323,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"v9rkceiE"} + data: {"id":"chatcmpl-CVgxGuhYP1UDMB0z0FHs1FIOt3DmR","object":"chat.completion.chunk","created":1761669322,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{"content":"Done"},"finish_reason":null}],"usage":null,"obfuscation":"Jtdb2O1z"} - data: {"id":"chatcmpl-CVfOVarszi9WEHgzSz5mFugUiAnQ8","object":"chat.completion.chunk","created":1761663323,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"D7s5lg"} + data: {"id":"chatcmpl-CVgxGuhYP1UDMB0z0FHs1FIOt3DmR","object":"chat.completion.chunk","created":1761669322,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[{"index":0,"delta":{},"finish_reason":"stop"}],"usage":null,"obfuscation":"Ys3ezs"} - data: {"id":"chatcmpl-CVfOVarszi9WEHgzSz5mFugUiAnQ8","object":"chat.completion.chunk","created":1761663323,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9851,"completion_tokens":4,"total_tokens":9855,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"IgZ45"} + data: {"id":"chatcmpl-CVgxGuhYP1UDMB0z0FHs1FIOt3DmR","object":"chat.completion.chunk","created":1761669322,"model":"gpt-5-2025-08-07","service_tier":"default","system_fingerprint":null,"choices":[],"usage":{"prompt_tokens":9600,"completion_tokens":4,"total_tokens":9604,"prompt_tokens_details":{"cached_tokens":8832,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"obfuscation":"HKVUN"} data: [DONE] @@ -215,4 +211,4 @@ interactions: - text/event-stream; charset=utf-8 status: 200 OK code: 200 - duration: 1.839327834s + duration: 3.780836583s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/bash_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/bash_tool.yaml index ceb43fbaf5164145c60f33995dae3310c47b1a00..b5b9a0ea97f4edc9420437842ccef0b927d91c21 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/bash_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/bash_tool.yaml @@ -24,21 +24,27 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":"Create"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":"Create"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":" test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":" test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt with"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":" hello bash using"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":" with"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":" bash"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":" hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":" bash"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663383-K2u41HGLLjFeR4fxktth","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":141,"completion_tokens":9,"total_tokens":150,"cost":0.00003195,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00002115,"upstream_inference_completions_cost":0.0000108},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":" using"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":" bash"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + + data: {"id":"gen-1761670365-Zes5c2usbsSNDZqJUKa8","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":141,"completion_tokens":9,"total_tokens":150,"cost":0.000024,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000141,"upstream_inference_completions_cost":0.0000099},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -47,15 +53,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 561.587917ms + duration: 1.288288584s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44272 + content_length: 43055 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -71,75 +77,65 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"bash:0","type":"function","function":{"name":"bash","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"command"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"bash:0","type":"function","function":{"name":"bash","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"command"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"echo"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" '"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"echo"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" bash"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" '"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"'"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" >"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" bash"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"'"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" >"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"description"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"description"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Create"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Create"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" with"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" content"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" '"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" with"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" bash"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" bash"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"'"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" content"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" using"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" echo"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" command"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663383-Nhd2fpQ2R9lvzsLdP1lJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663383,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10925,"completion_tokens":41,"total_tokens":10966,"cost":0.0018191,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017166,"upstream_inference_completions_cost":0.0001025},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761670365-bHmX4Bmvii8Yrw43wuBn","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10619,"completion_tokens":36,"total_tokens":10655,"cost":0.0017382,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016482,"upstream_inference_completions_cost":0.00009},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -148,100 +144,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.730197458s + duration: 2.794140666s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44581 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"bash:0\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"echo 'hello bash' > test.txt\\\", \\\"description\\\": \\\"Create test.txt file with content 'hello bash' using echo command\\\"}\",\"name\":\"bash\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"bash:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://openrouter.ai/api/v1/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"view:1","type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/bash"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663388-rFMClzH6ghhYHexL4web","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663388,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10985,"completion_tokens":32,"total_tokens":11017,"cost":0.0018326,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017526,"upstream_inference_completions_cost":0.00008},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} - - data: [DONE] - - headers: - Content-Type: - - text/event-stream - status: 200 OK - code: 200 - duration: 2.429436375s -- id: 3 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 44871 + content_length: 43343 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"bash:0\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"echo 'hello bash' > test.txt\\\", \\\"description\\\": \\\"Create test.txt file with content 'hello bash' using echo command\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"bash:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"view:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/bash_tool/test.txt\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|hello bash\\n\\n\",\"tool_call_id\":\"view:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"bash:0\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"echo 'hello bash' > test.txt\\\", \\\"description\\\": \\\"Create test.txt file with hello bash content\\\"}\",\"name\":\"bash\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"bash:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -257,21 +168,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":"Created"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":" test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":" successfully"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670369-juC60bptQ9kxe5rzoT9Y","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670369,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670369-juC60bptQ9kxe5rzoT9Y","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670369,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761670369-juC60bptQ9kxe5rzoT9Y","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670369,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663391-59dhAfLxJMLFpwkA3Mvj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663391,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11045,"completion_tokens":6,"total_tokens":11051,"cost":0.0018036,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017886,"upstream_inference_completions_cost":0.000015},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761670369-juC60bptQ9kxe5rzoT9Y","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761670369,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10674,"completion_tokens":2,"total_tokens":10676,"cost":0.0016862,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016812,"upstream_inference_completions_cost":0.000005},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -280,4 +183,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.649368375s + duration: 1.439740083s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/download_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/download_tool.yaml index d2bede44a22bbd800323ea757e02dd13abecf880..496c21caced69d37b20fc3194d767ed77c38fe9d 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/download_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/download_tool.yaml @@ -24,25 +24,27 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":"Download"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":"Download"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":" example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt from example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":" save"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":"-files"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":" example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":".online"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":"-"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":" from"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":"convert.com"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":" provided"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":" URL"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663393-oRK4e9XRMeWT1brxshNW","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":150,"completion_tokens":11,"total_tokens":161,"cost":0.0000357,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000225,"upstream_inference_completions_cost":0.0000132},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + + data: {"id":"gen-1761669416-XnOq3iDnpuzuT2hFao5y","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":150,"completion_tokens":9,"total_tokens":159,"cost":0.0000249,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.000015,"upstream_inference_completions_cost":0.0000099},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -51,15 +53,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 588.401292ms + duration: 791.584375ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44325 + content_length: 43108 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -75,65 +77,61 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"download:0","type":"function","function":{"name":"download","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"download","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"url"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"{\"url"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"https"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"https"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"://"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"://"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"example"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"example"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-files"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"-files"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".online"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":".online"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-con"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"-con"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"vert"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"vert"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".com"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":".com"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/document"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"/document"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"/"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"txt"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/example"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"/example"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":".txt"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"\","}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"_path"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"example"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"example"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".txt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":".txt"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.download:0","type":"function","function":{"name":"","arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663393-oL9KLq8KLZJXOx8nQlNd","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663393,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10933,"completion_tokens":35,"total_tokens":10968,"cost":0.0018089,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017214,"upstream_inference_completions_cost":0.0000875},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669416-BW3If5l8HzGbiMyjv9LD","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10315,"completion_tokens":36,"total_tokens":10351,"cost":0.006279,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.006189,"upstream_inference_completions_cost":0.00009},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -142,7 +140,7 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.322705709s + duration: 1.107660708s - id: 2 request: proto: HTTP/1.1 @@ -188,15 +186,15 @@ interactions: - text/plain; charset=UTF-8 status: 200 OK code: 200 - duration: 83.719041ms + duration: 577.568167ms - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44769 + content_length: 43572 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"download:0\",\"function\":{\"arguments\":\"{\\\"url\\\": \\\"https://example-files.online-convert.com/document/txt/example.txt\\\", \\\"file_path\\\": \\\"example.txt\\\"}\",\"name\":\"download\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"download:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.download:0\",\"function\":{\"arguments\":\"{\\\"url\\\": \\\"https://example-files.online-convert.com/document/txt/example.txt\\\", \\\"file_path\\\": \\\"example.txt\\\"}\",\"name\":\"download\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"functions.download:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -212,37 +210,49 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" Download"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":"ed"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" file"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" saved"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":" Saved"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" it"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" as"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":"2"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":"example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":"574"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":" bytes"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":" as"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":"example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" current"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":".txt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":" directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":"`."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663397-o1VB5sbI1hPVSJfCTBCy","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663397,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11021,"completion_tokens":14,"total_tokens":11035,"cost":0.0018092,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017742,"upstream_inference_completions_cost":0.000035},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669418-uuznEcLXJiPQILou0RHG","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669419,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10029,"completion_tokens":20,"total_tokens":10049,"cost":0.0060674,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0060174,"upstream_inference_completions_cost":0.00005},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -251,4 +261,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.046102958s + duration: 7.571912875s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/glob_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/glob_tool.yaml index 9f601e472856bb132c2a793dee00874d9bca406b..afe2a4272ee4dd198596c45a29a7f5c302a72917 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/glob_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/glob_tool.yaml @@ -24,31 +24,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":"Find"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":"Find"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" all"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":" all .go files"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" ."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":" in current directory using"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":" glob"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" files"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" current"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" using"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":" glob"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - - data: {"id":"gen-1761663413-P62iMlylSDbesyEeCNqy","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663413,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":139,"completion_tokens":11,"total_tokens":150,"cost":0.000026,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000139,"upstream_inference_completions_cost":0.0000121},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669427-uRptsm7WmJU0ZaWkoZ9X","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":139,"completion_tokens":11,"total_tokens":150,"cost":0.00003405,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00002085,"upstream_inference_completions_cost":0.0000132},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -57,15 +45,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 768.62575ms + duration: 766.656875ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44261 + content_length: 43044 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -81,25 +69,25 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"glob:0","type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":"functions.glob:0","index":0,"type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"{\"pattern"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"pattern"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"*."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"*."}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}]} - data: {"id":"gen-1761663413-wLM5cYxs4HXL1irYuTJb","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663414,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10923,"completion_tokens":15,"total_tokens":10938,"cost":0.0017529,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017154,"upstream_inference_completions_cost":0.0000375},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669427-d7vqRnAtRPty0J7Q1nv8","provider":"Chutes","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669427,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":9931,"completion_tokens":17,"total_tokens":9948,"cost":0.00390539,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00387309,"upstream_inference_completions_cost":0.0000323},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -108,15 +96,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.653864291s + duration: 3.776391417s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44516 + content_length: 43318 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"glob:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/glob_tool/main.go\",\"tool_call_id\":\"glob:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.glob:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/glob_tool/main.go\",\"tool_call_id\":\"functions.glob:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -132,15 +120,31 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663416-utSy43CYdElhR9nrXkBk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":" ."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":" file"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} + + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":"main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663416-utSy43CYdElhR9nrXkBk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663416,"choices":[{"index":0,"delta":{"role":"assistant","content":"main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663416-utSy43CYdElhR9nrXkBk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663416,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663416-utSy43CYdElhR9nrXkBk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":null} - data: {"id":"gen-1761663416-utSy43CYdElhR9nrXkBk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663416,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10972,"completion_tokens":3,"total_tokens":10975,"cost":0.0017523,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017448,"upstream_inference_completions_cost":0.0000075},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669431-QkpYN8O00LN6UnPiP5eG","provider":"BaseTen","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10356,"completion_tokens":12,"total_tokens":10368,"cost":0.0062436,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0062136,"upstream_inference_completions_cost":0.00003},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -149,4 +153,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.641015458s + duration: 523.052667ms diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/grep_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/grep_tool.yaml index b704c023d12c700ed823e170bdaff52cd4d8d506..e1864c6ea1a8709a221b1073074e6c29b97e8c00 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/grep_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/grep_tool.yaml @@ -24,17 +24,17 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663418-xOeNR41Tucke1vPV49JQ","provider":"Hyperbolic","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663418,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669432-27FSva2W9qyASe6gBAHL","provider":"SiliconFlow","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669432,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663418-xOeNR41Tucke1vPV49JQ","provider":"Hyperbolic","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663418,"choices":[{"index":0,"delta":{"role":"assistant","content":"Search"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669432-27FSva2W9qyASe6gBAHL","provider":"SiliconFlow","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669432,"choices":[{"index":0,"delta":{"role":"assistant","content":"Search"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663418-xOeNR41Tucke1vPV49JQ","provider":"Hyperbolic","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663418,"choices":[{"index":0,"delta":{"role":"assistant","content":" for package in Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669432-27FSva2W9qyASe6gBAHL","provider":"SiliconFlow","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669432,"choices":[{"index":0,"delta":{"role":"assistant","content":" for package in Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663418-xOeNR41Tucke1vPV49JQ","provider":"Hyperbolic","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663418,"choices":[{"index":0,"delta":{"role":"assistant","content":" files using grep"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669432-27FSva2W9qyASe6gBAHL","provider":"SiliconFlow","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669432,"choices":[{"index":0,"delta":{"role":"assistant","content":" files using grep"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663418-xOeNR41Tucke1vPV49JQ","provider":"Hyperbolic","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663418,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761669432-27FSva2W9qyASe6gBAHL","provider":"SiliconFlow","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669432,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663418-xOeNR41Tucke1vPV49JQ","provider":"Hyperbolic","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663418,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":140,"completion_tokens":9,"total_tokens":149,"cost":0.0000447,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.000042,"upstream_inference_completions_cost":0.0000027},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669432-27FSva2W9qyASe6gBAHL","provider":"SiliconFlow","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669432,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":140,"completion_tokens":9,"total_tokens":149,"cost":0.0000322,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000196,"upstream_inference_completions_cost":0.0000126},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -43,15 +43,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.378490084s + duration: 1.604201625s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44259 + content_length: 43042 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -67,37 +67,35 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"grep:0","type":"function","function":{"name":"grep","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":"functions.grep:0","index":0,"type":"function","function":{"name":"grep","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"{\"pattern"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"pattern"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"package"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"package"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\","}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"include"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"include"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \"*."}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"*."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}]} - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663418-bxggCbRtJR0SScqYQDnh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663419,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10924,"completion_tokens":22,"total_tokens":10946,"cost":0.001771,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001716,"upstream_inference_completions_cost":0.000055},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669432-r2twWcaHGH6d2XQXk8HB","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669433,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":9932,"completion_tokens":23,"total_tokens":9955,"cost":0.00553137,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0054626,"upstream_inference_completions_cost":0.00006877},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -106,15 +104,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.208485042s + duration: 2.260439792s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44592 + content_length: 43395 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"grep:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"package\\\", \\\"include\\\": \\\"*.go\\\"}\",\"name\":\"grep\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"grep:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.grep:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"package\\\", \\\"include\\\": \\\"*.go\\\"}\",\"name\":\"grep\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"functions.grep:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -130,65 +128,65 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":" match"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":" match"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":" `/"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":" `/"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"tmp"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"tmp"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"/c"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"/c"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"rush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"rush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"-test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"-test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"/Test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"/Test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"Coder"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"Coder"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"Agent"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"Agent"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"/open"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"/open"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"router"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"router"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"-k"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"-k"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"imi"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"imi"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"-k"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"-k"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"2"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"2"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"/g"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"/g"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"rep"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"rep"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"_tool"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"_tool"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"/main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"/main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663421-TexPcDeofler0LFhxcAo","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663421,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10999,"completion_tokens":28,"total_tokens":11027,"cost":0.001831,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001761,"upstream_inference_completions_cost":0.00007},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669436-B6RawSCCPNQiKPgiOQHv","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669436,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10693,"completion_tokens":28,"total_tokens":10721,"cost":0.0017626,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016926,"upstream_inference_completions_cost":0.00007},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -197,4 +195,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.675038416s + duration: 4.237480041s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/ls_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/ls_tool.yaml index fe8e25c96502a923ac8f950d7413f879a6a9922f..08f91baf8796b5825a52eb37932029daceeded58 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/ls_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/ls_tool.yaml @@ -24,17 +24,25 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663425-shUhmWj4YZ3iptJVZMNo","provider":"GMICloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663425-shUhmWj4YZ3iptJVZMNo","provider":"GMICloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":"List"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":"List"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663425-shUhmWj4YZ3iptJVZMNo","provider":"GMICloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":" files in current directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":" files"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663425-shUhmWj4YZ3iptJVZMNo","provider":"GMICloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":" using ls"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663425-shUhmWj4YZ3iptJVZMNo","provider":"GMICloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":" current"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663425-shUhmWj4YZ3iptJVZMNo","provider":"GMICloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":137,"completion_tokens":8,"total_tokens":145,"cost":0.00003255,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00002055,"upstream_inference_completions_cost":0.000012},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":" directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":" using"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":" ls"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + + data: {"id":"gen-1761669441-XvtAFoWGJNnwJ39Ct7gn","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":137,"completion_tokens":8,"total_tokens":145,"cost":0.0000225,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000137,"upstream_inference_completions_cost":0.0000088},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -43,15 +51,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.034993125s + duration: 596.613625ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44253 + content_length: 43036 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -67,55 +75,55 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"ls:0","type":"function","function":{"name":"ls","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"ls:0","type":"function","function":{"name":"ls","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ls"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ls"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663425-vH8IR8CCnpEjwXHY2vJ4","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663425,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10921,"completion_tokens":31,"total_tokens":10952,"cost":0.0017917,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017142,"upstream_inference_completions_cost":0.0000775},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669441-N6Yrr3ZtG35G4wbhwGg0","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669441,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10615,"completion_tokens":31,"total_tokens":10646,"cost":0.0017233,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016458,"upstream_inference_completions_cost":0.0000775},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -124,15 +132,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.97482525s + duration: 2.988043s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44574 + content_length: 43357 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"ls:0\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool\\\"}\",\"name\":\"ls\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"ls:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"ls:0\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool\\\"}\",\"name\":\"ls\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"ls:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -148,21 +156,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":".mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":".mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663428-jj9DmBi6oFr3DsfRB2k3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663428,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10998,"completion_tokens":6,"total_tokens":11004,"cost":0.0017754,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017604,"upstream_inference_completions_cost":0.000015},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669445-wN1HvfnzrTbXoIogC6gC","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669445,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10692,"completion_tokens":6,"total_tokens":10698,"cost":0.001707,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001692,"upstream_inference_completions_cost":0.000015},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -171,4 +179,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.837167792s + duration: 2.030201083s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/multiedit_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/multiedit_tool.yaml index c97888659f4bbc87fa927368e231ba6bc7c25f57..6521bf5b7a53f1c7a0628bab01e3d56575cf5138 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/multiedit_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/multiedit_tool.yaml @@ -24,37 +24,33 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"Use"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":"Update"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" mult"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" greeting"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"ied"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"it"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" update"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" with"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" greeting"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" mult"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":"ied"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" add"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":"it"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" comment"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":" comment"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - - data: {"id":"gen-1761663431-9w62WSqNusEUlRFYfGU3","provider":"AtlasCloud","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":160,"completion_tokens":14,"total_tokens":174,"cost":0.000045,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.000024,"upstream_inference_completions_cost":0.000021},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669448-FEqg80CuyXTh4cdr2OVe","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":160,"completion_tokens":12,"total_tokens":172,"cost":0.0000292,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.000016,"upstream_inference_completions_cost":0.0000132},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -63,15 +59,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 760.225333ms + duration: 1.200423958s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44339 + content_length: 43122 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -87,117 +83,63 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"I'll"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" help"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" you"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" use"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" mult"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"ied"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"it"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" make"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" those"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" changes"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" Let"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" me"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" first"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"view:0","type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" read"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" file"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" understand"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" its"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" current"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":" structure"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"view:0","type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/mult"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ied"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"it"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/mult"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ied"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"it"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663431-MYZTdU3cUXiSryra5lxS","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663431,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10945,"completion_tokens":61,"total_tokens":11006,"cost":0.0018811,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017286,"upstream_inference_completions_cost":0.0001525},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669448-b8wCPLzP6qH8OyIjWkMX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669448,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10639,"completion_tokens":34,"total_tokens":10673,"cost":0.0017452,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016602,"upstream_inference_completions_cost":0.000085},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -206,15 +148,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.22171175s + duration: 2.746319959s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44887 + content_length: 43532 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"I'll help you use multiedit to make those changes in main.go. Let me first read the file to understand its current structure.\",\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -230,245 +172,167 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"Now"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" I'll"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" use"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" mult"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"ied"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"it"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" make"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" both"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" changes"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"multiedit:1","type":"function","function":{"name":"multiedit","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" -"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" add"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" comment"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" above"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" fmt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":".Println"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" change"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" string"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" from"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" '"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" World"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"!'"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" '"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/mult"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ied"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"it"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":" Crush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"!"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":"'."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"multiedit:1","type":"function","function":{"name":"multiedit","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ed"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"its"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" ["},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\n\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"old"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_string"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"()"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" {\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\t"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"fmt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/mult"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".Println"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ied"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"(\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"it"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" World"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"!\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\")"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ed"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"its"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"new"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_string"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" [\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" "},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" {\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" "},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"()"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"old"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" {\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_string"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\t"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"//"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" G"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"reeting"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"()"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" {\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\t"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"fmt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\t"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".Println"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"fmt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"(\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".Println"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"(\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" Crush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"!\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" World"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\")"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"!\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}\n\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\")"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"]}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\",\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" "},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"new"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_string"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"()"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" {\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\t"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"//"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" G"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"reeting"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\t"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"fmt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".Println"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"(\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" Crush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"!\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\")"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" "},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" }\n"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"]}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663436-pqM8ceDA9NyOC8d6R9ca","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663436,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11074,"completion_tokens":127,"total_tokens":11201,"cost":0.0021235,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001806,"upstream_inference_completions_cost":0.0003175},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669451-xs0VI6ik9Sfu1CxSedkX","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669452,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10741,"completion_tokens":88,"total_tokens":10829,"cost":0.0019414,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017214,"upstream_inference_completions_cost":0.00022},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -477,15 +341,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.869071792s + duration: 2.760361458s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45638 + content_length: 44115 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"I'll help you use multiedit to make those changes in main.go. Let me first read the file to understand its current structure.\",\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"},{\"content\":\"Now I'll use multiedit to make both changes - add the comment above fmt.Println and change the string from 'Hello, World!' to 'Hello, Crush!'.\",\"tool_calls\":[{\"id\":\"multiedit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\", \\\"edits\\\": [\\n {\\n \\\"old_string\\\": \\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\n \\\"new_string\\\": \\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\"\\n }\\n]}\",\"name\":\"multiedit\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"multiedit:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"multiedit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\", \\\"edits\\\": [\\n\\n{\\\"old_string\\\": \\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\", \\\"new_string\\\": \\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\"}\\n\\n]}\",\"name\":\"multiedit\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"multiedit:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -501,180 +365,41 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"view:2","type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/mult"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"ied"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"it"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663445-INa2ySPUMI0b2e6tSCLD","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663445,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11253,"completion_tokens":34,"total_tokens":11287,"cost":0.0018832,"is_byok":false,"prompt_tokens_details":{"cached_tokens":11008,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017982,"upstream_inference_completions_cost":0.000085},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} - - data: [DONE] - - headers: - Content-Type: - - text/event-stream - status: 200 OK - code: 200 - duration: 2.667102375s -- id: 4 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 46070 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"I'll help you use multiedit to make those changes in main.go. Let me first read the file to understand its current structure.\",\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"},{\"content\":\"Now I'll use multiedit to make both changes - add the comment above fmt.Println and change the string from 'Hello, World!' to 'Hello, Crush!'.\",\"tool_calls\":[{\"id\":\"multiedit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\", \\\"edits\\\": [\\n {\\n \\\"old_string\\\": \\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\n \\\"new_string\\\": \\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\"\\n }\\n]}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"multiedit:1\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"view:2\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\t// Greeting\\n 7|\\tfmt.Println(\\\"Hello, Crush!\\\")\\n 8|}\\n\\n\",\"tool_call_id\":\"view:2\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://openrouter.ai/api/v1/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" I've"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" successfully"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" used"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" mult"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"ied"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"it"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":":\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" Add"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" comment"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"//"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" G"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"reeting"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" above"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" fmt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":".Println"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" line"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"2"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" Changed"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" Changed"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" \""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" greeting"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" added"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" World"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"!\""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" comment"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" above"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" \""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" fmt"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":".Println"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":" Crush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":" line"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":"!\""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663449-ZnJKoA53WR0VmlrSmR14","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663449,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11364,"completion_tokens":41,"total_tokens":11405,"cost":0.0019673,"is_byok":false,"prompt_tokens_details":{"cached_tokens":11008,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0018648,"upstream_inference_completions_cost":0.0001025},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669459-ilh3kL6vqUTD8VrtLdXW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669459,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10881,"completion_tokens":16,"total_tokens":10897,"cost":0.0018454,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0018054,"upstream_inference_completions_cost":0.00004},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -683,4 +408,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.136257792s + duration: 2.19689125s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls.yaml index a12b2f5d8715dcb272bff189ac74117138f3e2c2..7c40a68db4bdd2c46105ecdf0934e017097fd28c 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls.yaml @@ -24,23 +24,23 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":"Run"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":"Find"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":" glob"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":" .go files and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":" and ls in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":" list directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":" parallel to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":" find .go files"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":" parallel"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":" and list directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663476-aqolUhapwap8EExXBGXr","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":156,"completion_tokens":10,"total_tokens":166,"cost":0.00003284,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00002184,"upstream_inference_completions_cost":0.000011},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669482-ZMZPqw2idrUAmxFQ2dwz","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":156,"completion_tokens":15,"total_tokens":171,"cost":0.0000414,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000234,"upstream_inference_completions_cost":0.000018},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -49,15 +49,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 565.017292ms + duration: 603.669334ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44350 + content_length: 43133 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -73,31 +73,37 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"glob:0","type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"glob:0","type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"pattern"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"pattern"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"**"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"*."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/*."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"id":"ls:1","type":"function","function":{"name":"ls","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"id":"ls:1","type":"function","function":{"name":"ls","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"function":{"arguments":"{}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"function":{"arguments":"path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663476-9gkVPtddVeks0FExEVxh","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663476,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10941,"completion_tokens":25,"total_tokens":10966,"cost":0.0017887,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017262,"upstream_inference_completions_cost":0.0000625},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"function":{"arguments":" \".\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":1,"function":{"arguments":"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + + data: {"id":"gen-1761669482-NH5f0FastyHd5f2l4Qje","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669482,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10635,"completion_tokens":28,"total_tokens":10663,"cost":0.0017278,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016578,"upstream_inference_completions_cost":0.00007},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -106,15 +112,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.627963583s + duration: 3.082027084s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44844 + content_length: 43639 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"glob:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"ls:1\",\"function\":{\"arguments\":\"{}\",\"name\":\"ls\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls/main.go\",\"tool_call_id\":\"glob:0\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"ls:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"glob:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\": \\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"ls:1\",\"function\":{\"arguments\":\"{\\\"path\\\": \\\".\\\"}\",\"name\":\"ls\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls/main.go\",\"tool_call_id\":\"glob:0\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"ls:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -130,89 +136,51 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" ."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" file"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" `/"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"tmp"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"/c"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"rush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"-test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"/Test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"Coder"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"Agent"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"/open"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"router"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"-k"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"imi"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"-k"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"2"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"/"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"parallel"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" ."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"_tool"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"_calls"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" file"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"/main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"`\n\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":"main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"Current"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":"`\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" contains"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":"Current"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":":\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" directory"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"-"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" contains"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":".mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":".mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":"-"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663479-7ZznDOV7P1IGPLvbL5oJ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663479,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11044,"completion_tokens":40,"total_tokens":11084,"cost":0.001888,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001788,"upstream_inference_completions_cost":0.0001},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669486-KZxax4J3DFYRx2z745fV","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669486,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10741,"completion_tokens":21,"total_tokens":10762,"cost":0.0017739,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017214,"upstream_inference_completions_cost":0.0000525},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -221,4 +189,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 3.089859s + duration: 2.193468458s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/read_a_file.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/read_a_file.yaml index 251369050607cc121612c754522c93ebf0f32187..61a86253b825898659e1975f8b2d5afcd0ca77e6 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/read_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/read_a_file.yaml @@ -24,15 +24,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663339-Q5rfOJKqq0tmpcRCQk9u","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663339-Q5rfOJKqq0tmpcRCQk9u","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":"Read"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":"Read"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663339-Q5rfOJKqq0tmpcRCQk9u","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":" the go mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":" the"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663339-Q5rfOJKqq0tmpcRCQk9u","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":" go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663339-Q5rfOJKqq0tmpcRCQk9u","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":131,"completion_tokens":5,"total_tokens":136,"cost":0.00002565,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00001965,"upstream_inference_completions_cost":0.000006},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":" mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + + data: {"id":"gen-1761669358-yFzArE7XzhfY3poXZFFV","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669358,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":131,"completion_tokens":5,"total_tokens":136,"cost":0.0000186,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000131,"upstream_inference_completions_cost":0.0000055},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -41,15 +45,66 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 417.198333ms + duration: 1.60030675s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44223 + content_length: 43006 + host: "" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Read the go mod\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + headers: + Accept: + - application/json + Content-Type: + - application/json + User-Agent: + - OpenAI/Go 2.7.1 + url: https://openrouter.ai/api/v1/chat/completions + method: POST + response: + proto: HTTP/2.0 + proto_major: 2 + proto_minor: 0 + content_length: -1 + body: |+ + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.glob:0","type":"function","function":{"name":"glob","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"{\""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"pattern"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"\":\""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":".mod"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":""} + + data: {"id":"gen-1761669358-dBxE7RWWXINZ4gtey4AG","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669360,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":9993,"completion_tokens":17,"total_tokens":10010,"cost":0.0040312,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0039972,"upstream_inference_completions_cost":0.000034},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + + data: [DONE] + + headers: + Content-Type: + - text/event-stream + status: 200 OK + code: 200 + duration: 8.601969125s +- id: 2 + request: + proto: HTTP/1.1 + proto_major: 1 + proto_minor: 1 + content_length: 43283 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Read the go mod\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.glob:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"go.mod\\\"}\",\"name\":\"glob\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file/go.mod\",\"tool_call_id\":\"functions.glob:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -65,59 +120,59 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":"functions.view:0","index":0,"type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.view:1","type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"{\"file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"{\""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_path"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"_path"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \"/"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"\":\"/"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"tmp"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"tmp"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/c"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"/c"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"rush"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"rush"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"-test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"-test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/Test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"/Test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Coder"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"Coder"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Agent"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"Agent"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/open"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"/open"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"router"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"router"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"-k"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"-k"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"imi"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"imi"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"-k"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"-k"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"2"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"2"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/read"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"/read"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_a"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"_a"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"_file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"/go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":".mod"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":".mod"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":null,"type":"function","function":{"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}]} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663339-KKkHwVXiwKQWEuo7KYjz","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663339,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10186,"completion_tokens":34,"total_tokens":10220,"cost":0.00570396,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0056023,"upstream_inference_completions_cost":0.00010166},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669368-tNaRnCLXQwycRpESgIZg","provider":"SiliconFlow","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669368,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10043,"completion_tokens":33,"total_tokens":10076,"cost":0.0040832,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0040172,"upstream_inference_completions_cost":0.000066},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -126,15 +181,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.626690084s -- id: 2 + duration: 5.031207542s +- id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44578 + content_length: 43637 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\n\",\"tool_call_id\":\"functions.view:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.glob:0\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"go.mod\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file/go.mod\",\"tool_call_id\":\"functions.glob:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"functions.view:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\n\",\"tool_call_id\":\"functions.view:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -150,61 +205,47 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"Your"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" module"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" file"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" shows"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":":\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"-"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" Module"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" name"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":".mod"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":" contents"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":":\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"```\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":".com"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"module"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"/test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":" example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"project"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":".com"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"`\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"/test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"-"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"project"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"\n\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" version"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"1"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"23"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"23"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":"```"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663341-xMEIoQ0S3Q4DhMHyWxvZ","provider":"Novita","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663357,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10993,"completion_tokens":26,"total_tokens":11019,"cost":0.0066608,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0065958,"upstream_inference_completions_cost":0.000065},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669376-QdTa7GnDNJ3y365A6tZF","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669376,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10051,"completion_tokens":19,"total_tokens":10070,"cost":0.0060781,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0060306,"upstream_inference_completions_cost":0.0000475},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -213,4 +254,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 10.091337834s + duration: 4.090280667s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/simple_test.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/simple_test.yaml index aa9d404950f302c007afcca11d4e34ad8c5d3c05..76e99e43dad4b56b24a6aff2da71ea50bdb1246c 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/simple_test.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/simple_test.yaml @@ -24,13 +24,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663336-chT2pUxbeiyfxOQHAMzY","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669340-MyVEOWu79L01Iq5ZcGwR","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669340,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663336-chT2pUxbeiyfxOQHAMzY","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669340-MyVEOWu79L01Iq5ZcGwR","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669340,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663336-chT2pUxbeiyfxOQHAMzY","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669340-MyVEOWu79L01Iq5ZcGwR","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669340,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663336-chT2pUxbeiyfxOQHAMzY","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":132,"completion_tokens":1,"total_tokens":133,"cost":0.0000213,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000198,"upstream_inference_completions_cost":0.0000015},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669340-MyVEOWu79L01Iq5ZcGwR","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669340,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":128,"completion_tokens":2,"total_tokens":130,"cost":0.000015,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000128,"upstream_inference_completions_cost":0.0000022},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -39,15 +39,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.08956625s + duration: 1.533934708s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44213 + content_length: 42996 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Hello\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"Hello\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -63,13 +63,11 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663336-R4FtmvJiS0oJyC4uob9R","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669341-Zc9tGLJYAXtornTSoKfX","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669356,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663336-R4FtmvJiS0oJyC4uob9R","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669341-Zc9tGLJYAXtornTSoKfX","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669356,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello!"},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663336-R4FtmvJiS0oJyC4uob9R","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - - data: {"id":"gen-1761663336-R4FtmvJiS0oJyC4uob9R","provider":"AtlasCloud","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663336,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10182,"completion_tokens":2,"total_tokens":10184,"cost":0.0061142,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0061092,"upstream_inference_completions_cost":0.000005},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669341-Zc9tGLJYAXtornTSoKfX","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669356,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":9999,"completion_tokens":2,"total_tokens":10001,"cost":0.0050035,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0049995,"upstream_inference_completions_cost":0.000004},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -78,4 +76,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.850610375s + duration: 10.389440792s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool.yaml index 2cf87b4f85f2f733dd438de436232f014b9ec2cd..b19fd09ed3f96b92f132e30a9669c0a1d2d2e7c4 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool.yaml @@ -24,19 +24,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":"Search"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":"Search"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":" for func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":" for func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":" main in Go repositories"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":" main in Go repositories"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":" using Sourcegraph"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":" using Source"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":""} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":"graph"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663453-yVFjIcfB2F5uGyiOPkS6","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":144,"completion_tokens":10,"total_tokens":154,"cost":0.0000366,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000216,"upstream_inference_completions_cost":0.000015},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + + data: {"id":"gen-1761669462-aRmFfGfAl74m2ZDSlhef","provider":"Google","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":140,"completion_tokens":11,"total_tokens":151,"cost":0.0000342,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.000021,"upstream_inference_completions_cost":0.0000132},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -45,15 +47,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 847.554708ms + duration: 656.152375ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44273 + content_length: 43056 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -69,55 +71,33 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"sourcegraph:0","type":"function","function":{"name":"sourcegraph","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"query"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"lang"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"sourcegraph:0","type":"function","function":{"name":"sourcegraph","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"query"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" type"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"lang"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"symbol"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"count"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" "},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"10"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663453-jfWuyf36lvMylU4Ys36z","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663453,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10924,"completion_tokens":31,"total_tokens":10955,"cost":0.0017935,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001716,"upstream_inference_completions_cost":0.0000775},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669462-mstnn1JiBNCjDVANC21C","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669462,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10618,"completion_tokens":20,"total_tokens":10638,"cost":0.0016976,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016476,"upstream_inference_completions_cost":0.00005},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -126,15 +106,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.743922458s + duration: 2.864345791s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 440 + content_length: 424 host: sourcegraph.com - body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"lang:go \"func main\" type:symbol"}}' + body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"lang:go func main"}}' headers: Content-Type: - application/json @@ -148,21 +128,21 @@ interactions: proto_minor: 0 content_length: -1 uncompressed: true - body: '{"data":{"search":{"results":{"matchCount":0,"limitHit":false,"resultCount":0,"approximateResultCount":"0","missing":[],"timedout":[],"indexUnavailable":false,"results":[]}}}}' + body: '{"data":{"search":{"results":{"matchCount":30,"limitHit":true,"resultCount":30,"approximateResultCount":"30+","missing":[],"timedout":[],"indexUnavailable":false,"results":[{"__typename":"FileMatch","repository":{"name":"github.com/jcalabro/uscope"},"file":{"path":"assets/gobacktrace/main.go","url":"/github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc FuncE() {\n\tfmt.Println(\"FuncE\")\n}\n\nfunc FuncD() {\n\tFuncE()\n\tfmt.Println(\"FuncD\")\n}\n\nfunc FuncC() {\n\tFuncD()\n\tfmt.Println(\"FuncC\")\n}\n\nfunc FuncB() {\n\tFuncC()\n\tfmt.Println(\"FuncB\")\n}\n\nfunc FuncA() {\n\tFuncB()\n\tfmt.Println(\"FuncA\")\n}\n\nfunc FuncF() {\n\tfmt.Println(\"FuncF\")\n\tFuncE()\n}\n\nfunc main() {\n\tFuncA()\n\tFuncB()\n\tFuncC()\n\tFuncD()\n\tFuncE()\n\tFuncF()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":33,"offsetAndLengths":[[0,9]]},{"preview":"\tFuncA()","lineNumber":34,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncB()","lineNumber":35,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncC()","lineNumber":36,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncD()","lineNumber":37,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncE()","lineNumber":38,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncF()","lineNumber":39,"offsetAndLengths":[[1,4]]},{"preview":"func FuncE() {","lineNumber":4,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tfmt.Println(\"FuncE\")","lineNumber":5,"offsetAndLengths":[[14,4]]},{"preview":"func FuncD() {","lineNumber":8,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncE()","lineNumber":9,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncD\")","lineNumber":10,"offsetAndLengths":[[14,4]]},{"preview":"func FuncC() {","lineNumber":13,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncD()","lineNumber":14,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncC\")","lineNumber":15,"offsetAndLengths":[[14,4]]},{"preview":"func FuncB() {","lineNumber":18,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncC()","lineNumber":19,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncB\")","lineNumber":20,"offsetAndLengths":[[14,4]]},{"preview":"func FuncA() {","lineNumber":23,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncB()","lineNumber":24,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncA\")","lineNumber":25,"offsetAndLengths":[[14,4]]},{"preview":"func FuncF() {","lineNumber":28,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tfmt.Println(\"FuncF\")","lineNumber":29,"offsetAndLengths":[[14,4]]},{"preview":"\tFuncE()","lineNumber":30,"offsetAndLengths":[[1,4]]}]}]}}}}' headers: Content-Type: - application/json status: 200 OK code: 200 - duration: 1.338321167s + duration: 1.461836208s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44638 + content_length: 50817 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"sourcegraph:0\",\"function\":{\"arguments\":\"{\\\"query\\\": \\\"lang:go \\\\\\\"func main\\\\\\\" type:symbol\\\", \\\"count\\\": 10}\",\"name\":\"sourcegraph\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 0 matches across 0 results\\n\\nNo results found. Try a different query.\\n\",\"tool_call_id\":\"sourcegraph:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"sourcegraph:0\",\"function\":{\"arguments\":\"{\\\"query\\\": \\\"lang:go func main\\\"}\",\"name\":\"sourcegraph\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/jcalabro/uscope/assets/gobacktrace/main.go\\n\\nURL: /github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go\\n\\n```\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| func main() {\\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| \\tFuncA()\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncB()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncC()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncD()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncE()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncF()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func FuncE() {\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\tfmt.Println(\\\"FuncE\\\")\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| func FuncD() {\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| \\tFuncE()\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tfmt.Println(\\\"FuncD\\\")\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n```\\n\\n```\\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| func FuncC() {\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n```\\n\\n```\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| \\tFuncD()\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n```\\n\\n```\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tfmt.Println(\\\"FuncC\\\")\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n```\\n\\n```\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| func FuncB() {\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n```\\n\\n```\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| \\tFuncC()\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n```\\n\\n```\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tfmt.Println(\\\"FuncB\\\")\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n```\\n\\n```\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| func FuncA() {\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n```\\n\\n```\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| \\tFuncB()\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n```\\n\\n```\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tfmt.Println(\\\"FuncA\\\")\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n```\\n\\n```\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| func FuncF() {\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n```\\n\\n```\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| \\tfmt.Println(\\\"FuncF\\\")\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n```\\n\\n```\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tFuncE()\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n```\\n\\n\",\"tool_call_id\":\"sourcegraph:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -178,222 +158,169 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"sourcegraph:1","type":"function","function":{"name":"sourcegraph","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"query"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"30"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" matches"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" across"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"lang"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"30"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" func"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" The"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" first"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" result"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"count"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" shows"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" a"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" "},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" simple"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"10"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" program"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" with"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663459-v1iTKywzmwtl7lQ70Wy3","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663459,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10996,"completion_tokens":26,"total_tokens":11022,"cost":0.0018242,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017592,"upstream_inference_completions_cost":0.000065},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" a"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: [DONE] + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - headers: - Content-Type: - - text/event-stream - status: 200 OK - code: 200 - duration: 2.274058792s -- id: 4 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 424 - host: sourcegraph.com - body: '{"query":"query Search($query: String!) { search(query: $query, version: V2, patternType: keyword ) { results { matchCount, limitHit, resultCount, approximateResultCount, missing { name }, timedout { name }, indexUnavailable, results { __typename, ... on FileMatch { repository { name }, file { path, url, content }, lineMatches { preview, lineNumber, offsetAndLengths } } } } } }","variables":{"query":"lang:go func main"}}' - headers: - Content-Type: - - application/json - User-Agent: - - crush/1.0 - url: https://sourcegraph.com/.api/graphql - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - uncompressed: true - body: '{"data":{"search":{"results":{"matchCount":30,"limitHit":true,"resultCount":30,"approximateResultCount":"30+","missing":[],"timedout":[],"indexUnavailable":false,"results":[{"__typename":"FileMatch","repository":{"name":"github.com/jcalabro/uscope"},"file":{"path":"assets/gobacktrace/main.go","url":"/github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go","content":"package main\n\nimport \"fmt\"\n\nfunc FuncE() {\n\tfmt.Println(\"FuncE\")\n}\n\nfunc FuncD() {\n\tFuncE()\n\tfmt.Println(\"FuncD\")\n}\n\nfunc FuncC() {\n\tFuncD()\n\tfmt.Println(\"FuncC\")\n}\n\nfunc FuncB() {\n\tFuncC()\n\tfmt.Println(\"FuncB\")\n}\n\nfunc FuncA() {\n\tFuncB()\n\tfmt.Println(\"FuncA\")\n}\n\nfunc FuncF() {\n\tfmt.Println(\"FuncF\")\n\tFuncE()\n}\n\nfunc main() {\n\tFuncA()\n\tFuncB()\n\tFuncC()\n\tFuncD()\n\tFuncE()\n\tFuncF()\n}\n"},"lineMatches":[{"preview":"func main() {","lineNumber":33,"offsetAndLengths":[[0,9]]},{"preview":"\tFuncA()","lineNumber":34,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncB()","lineNumber":35,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncC()","lineNumber":36,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncD()","lineNumber":37,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncE()","lineNumber":38,"offsetAndLengths":[[1,4]]},{"preview":"\tFuncF()","lineNumber":39,"offsetAndLengths":[[1,4]]},{"preview":"func FuncE() {","lineNumber":4,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tfmt.Println(\"FuncE\")","lineNumber":5,"offsetAndLengths":[[14,4]]},{"preview":"func FuncD() {","lineNumber":8,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncE()","lineNumber":9,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncD\")","lineNumber":10,"offsetAndLengths":[[14,4]]},{"preview":"func FuncC() {","lineNumber":13,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncD()","lineNumber":14,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncC\")","lineNumber":15,"offsetAndLengths":[[14,4]]},{"preview":"func FuncB() {","lineNumber":18,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncC()","lineNumber":19,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncB\")","lineNumber":20,"offsetAndLengths":[[14,4]]},{"preview":"func FuncA() {","lineNumber":23,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tFuncB()","lineNumber":24,"offsetAndLengths":[[1,4]]},{"preview":"\tfmt.Println(\"FuncA\")","lineNumber":25,"offsetAndLengths":[[14,4]]},{"preview":"func FuncF() {","lineNumber":28,"offsetAndLengths":[[0,4],[5,4]]},{"preview":"\tfmt.Println(\"FuncF\")","lineNumber":29,"offsetAndLengths":[[14,4]]},{"preview":"\tFuncE()","lineNumber":30,"offsetAndLengths":[[1,4]]}]}]}}}}' - headers: - Content-Type: - - application/json - status: 200 OK - code: 200 - duration: 1.226600625s -- id: 5 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 52414 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"sourcegraph:0\",\"function\":{\"arguments\":\"{\\\"query\\\": \\\"lang:go \\\\\\\"func main\\\\\\\" type:symbol\\\", \\\"count\\\": 10}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 0 matches across 0 results\\n\\nNo results found. Try a different query.\\n\",\"tool_call_id\":\"sourcegraph:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"sourcegraph:1\",\"function\":{\"arguments\":\"{\\\"query\\\": \\\"lang:go func main\\\", \\\"count\\\": 10}\",\"name\":\"sourcegraph\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/jcalabro/uscope/assets/gobacktrace/main.go\\n\\nURL: /github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go\\n\\n```\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| func main() {\\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| \\tFuncA()\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncB()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncC()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncD()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncE()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncF()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func FuncE() {\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\tfmt.Println(\\\"FuncE\\\")\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| func FuncD() {\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| \\tFuncE()\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tfmt.Println(\\\"FuncD\\\")\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n```\\n\\n```\\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| func FuncC() {\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n```\\n\\n```\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| \\tFuncD()\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n```\\n\\n```\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tfmt.Println(\\\"FuncC\\\")\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n```\\n\\n```\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| func FuncB() {\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n```\\n\\n```\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| \\tFuncC()\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n```\\n\\n```\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tfmt.Println(\\\"FuncB\\\")\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n```\\n\\n```\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| func FuncA() {\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n```\\n\\n```\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| \\tFuncB()\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n```\\n\\n```\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tfmt.Println(\\\"FuncA\\\")\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n```\\n\\n```\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| func FuncF() {\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n```\\n\\n```\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| \\tfmt.Println(\\\"FuncF\\\")\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n```\\n\\n```\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tFuncE()\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n```\\n\\n\",\"tool_call_id\":\"sourcegraph:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://openrouter.ai/api/v1/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" function"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"Found"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" that"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" "},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" calls"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"30"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" several"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" matches"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" other"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" across"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" functions"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" ("},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"Func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"A"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" The"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" through"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" results"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" Func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" show"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"F"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" various"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":")"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" sequence"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" at"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"()`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" github"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" implementations"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":".com"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"/j"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" including"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"cal"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" a"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"ab"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" simple"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"ro"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" example"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"/usc"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" from"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"ope"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" github"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"/assets"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":".com"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"/g"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"/j"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"ob"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"cal"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"ack"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"ab"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"trace"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"ro"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"/main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"/usc"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"ope"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"/assets"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"34"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"/g"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":".\n\n"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"ob"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"The"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"ack"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" results"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"trace"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" show"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"/main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" various"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" Go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" programs"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"34"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" containing"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" that"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" `"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" calls"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" a"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" series"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"()`"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" of"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" declarations"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" functions"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":","},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" ("},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" demonstrating"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"Func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" different"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"A"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" implementations"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" through"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" and"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" Func"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" use"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"F"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" cases"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":")"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" across"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" in"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" public"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":" sequence"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663463-k7pNpdr21v43xFaLC9nw","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663463,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":13826,"completion_tokens":57,"total_tokens":13883,"cost":0.0035997,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0034572,"upstream_inference_completions_cost":0.0001425},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669467-B5Et5HxcxpWl5Q1k7Yaj","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669467,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":13442,"completion_tokens":80,"total_tokens":13522,"cost":0.003542,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.003342,"upstream_inference_completions_cost":0.0002},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -402,4 +329,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.087341583s + duration: 2.38255575s diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/update_a_file.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/update_a_file.yaml index b7171fd5126a1b82ef4796a2d775de6035e0756c..f2b2cd82acab58daf68ce1b06138707c7d4733f5 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/update_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/update_a_file.yaml @@ -24,27 +24,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":"Update"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":"Update"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":" main.go to print"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":" hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":" from crush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":" print"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":""} - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":" hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":" from"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":" crush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - - data: {"id":"gen-1761663365-WDezLwg1xKSFVAbuGxk2","provider":"Parasail","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":141,"completion_tokens":9,"total_tokens":150,"cost":0.000024,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000141,"upstream_inference_completions_cost":0.0000099},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669380-2UpCxj2VmJCteRG6TEq1","provider":"Novita","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669380,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":145,"completion_tokens":8,"total_tokens":153,"cost":0.00003375,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.00002175,"upstream_inference_completions_cost":0.000012},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -53,15 +45,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 625.86275ms + duration: 775.037584ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44279 + content_length: 43062 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -77,61 +69,11 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"view:0","type":"function","function":{"name":"view","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669380-GGOoK6lf1mJAdat1CPkR","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669396,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/update"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669380-GGOoK6lf1mJAdat1CPkR","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669396,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"functions.view:0","function":{"arguments":"{\"file_path\": \"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\"}","name":"view"},"type":"function"}]},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}]} - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_a"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663365-CI7krv4bTJLhkryU0fZf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663365,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10925,"completion_tokens":33,"total_tokens":10958,"cost":0.0017991,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017166,"upstream_inference_completions_cost":0.0000825},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669380-GGOoK6lf1mJAdat1CPkR","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669396,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10814,"completion_tokens":24,"total_tokens":10838,"cost":0.005455,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.005407,"upstream_inference_completions_cost":0.000048},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -140,15 +82,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.805592875s + duration: 10.071588375s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44688 + content_length: 43491 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"functions.view:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -164,121 +106,119 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"edit:1","type":"function","function":{"name":"edit","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":"functions.edit:1","index":0,"type":"function","function":{"name":"edit","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"{\"file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_path"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \"/"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"tmp"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/c"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"rush"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"-test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/Test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Coder"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Agent"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/open"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"router"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"-k"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"imi"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"-k"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"2"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/update"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/update"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_a"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_a"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_file"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"/main"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":".go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\","}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"old"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"old"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_string"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_string"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \"\\"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"tf"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tf"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"mt"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"mt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":".Println"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".Println"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"(\\\""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"(\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Hello"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":","}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" World"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" World"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"!\\"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"!\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\")"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\")"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\","}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"new"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"new"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"_string"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_string"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \"\\"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"tf"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tf"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"mt"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"mt"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":".Println"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".Println"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"(\\\""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"(\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Hello"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Hello"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" from"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" from"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" Crush"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" Crush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"!\\"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"!\\"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\")"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\")"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}]} - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663369-d3vAMI4kBtQckNCf3UGT","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663369,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11026,"completion_tokens":63,"total_tokens":11089,"cost":0.0019347,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017772,"upstream_inference_completions_cost":0.0001575},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669398-5KPckpRaoP09EBNe3Eya","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669398,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10034,"completion_tokens":64,"total_tokens":10098,"cost":0.00571006,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0055187,"upstream_inference_completions_cost":0.00019136},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -287,15 +227,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.971172916s + duration: 1.241644s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45187 + content_length: 44010 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"edit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\", \\\"new_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello from Crush!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"edit:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"functions.view:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"functions.edit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\", \\\"new_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello from Crush!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"functions.edit:1\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -311,51 +251,47 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"bash:2","type":"function","function":{"name":"bash","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"command"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":"functions.bash:2","index":0,"type":"function","function":{"name":"bash","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"{\"command"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" run"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" main"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" run"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" main"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":".go"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\","}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"description"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"description"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\":"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" \""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" the"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"Test"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" updated"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" the"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" Go"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" updated"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" program"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":" program"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"id":null,"index":0,"type":"function","function":{"name":null,"arguments":"\"}"}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}]} - data: {"id":"gen-1761663376-pWtbZCKmAfgcgkh2ozqk","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663376,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11136,"completion_tokens":29,"total_tokens":11165,"cost":0.0018005,"is_byok":false,"prompt_tokens_details":{"cached_tokens":11008,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001728,"upstream_inference_completions_cost":0.0000725},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669402-WLvQs62KrCEwGHxfjBXF","provider":"Parasail","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669402,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10144,"completion_tokens":29,"total_tokens":10173,"cost":0.00566591,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0055792,"upstream_inference_completions_cost":0.00008671},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -364,15 +300,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.950160167s + duration: 1.02377225s - id: 4 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45532 + content_length: 44372 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"view:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"edit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\", \\\"new_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello from Crush!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"edit:1\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"bash:2\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"go run main.go\\\", \\\"description\\\": \\\"Test the updated Go program\\\"}\",\"name\":\"bash\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"Hello from Crush!\\n\\n\\n/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\",\"tool_call_id\":\"bash:2\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"functions.view:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"functions.view:0\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"functions.edit:1\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\\", \\\"old_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\", \\\"new_string\\\": \\\"\\\\tfmt.Println(\\\\\\\"Hello from Crush!\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"functions.edit:1\",\"role\":\"tool\"},{\"tool_calls\":[{\"id\":\"functions.bash:2\",\"function\":{\"arguments\":\"{\\\"command\\\": \\\"go run main.go\\\", \\\"description\\\": \\\"Test the updated program\\\"}\",\"name\":\"bash\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"Hello from Crush!\\n\\n\\n/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/update_a_file\",\"tool_call_id\":\"functions.bash:2\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -388,39 +324,11 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":"."},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" Updated"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" main"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":".go"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":":"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":"6"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" to"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" print"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" \""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" from"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":" Crush"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":"!\""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669404-C8WMjrGU1AuJJQmfQV7X","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669404,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669404-C8WMjrGU1AuJJQmfQV7X","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669404,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663380-X3SWJMK8c3DQfq831YuQ","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663381,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11209,"completion_tokens":15,"total_tokens":11224,"cost":0.0018093,"is_byok":false,"prompt_tokens_details":{"cached_tokens":11008,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017718,"upstream_inference_completions_cost":0.0000375},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669404-C8WMjrGU1AuJJQmfQV7X","provider":"DeepInfra","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669404,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10308,"completion_tokens":1,"total_tokens":10309,"cost":0.005156,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.005154,"upstream_inference_completions_cost":0.000002},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -429,4 +337,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.006659458s + duration: 816.996125ms diff --git a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/write_tool.yaml b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/write_tool.yaml index 11f7705c1b65ee0eaf49a09034e2c862430c2b79..63658d4a3e5987b1a7d2e29ff6b6c28f0f94d840 100644 --- a/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/write_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/openrouter-kimi-k2/write_tool.yaml @@ -24,19 +24,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":"Create"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":"Create"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":" config.json with name"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":" config.json with test"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":" and version data"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":" data"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}]} - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}]} - data: {"id":"gen-1761663468-FdSBO5s1bywNMrHIsGKy","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761663468,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":155,"completion_tokens":9,"total_tokens":164,"cost":0.0000316,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000217,"upstream_inference_completions_cost":0.0000099},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669474-LAUlgDUUs9eehWLok40t","provider":"DeepInfra","model":"qwen/qwen3-next-80b-a3b-instruct","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":155,"completion_tokens":7,"total_tokens":162,"cost":0.0000294,"is_byok":false,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0000217,"upstream_inference_completions_cost":0.0000077},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -45,15 +45,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 587.740334ms + duration: 804.668542ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44316 + content_length: 43099 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -69,101 +69,101 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"write:0","type":"function","function":{"name":"write","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"id":"write:0","type":"function","function":{"name":"write","arguments":""}}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"{\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"file"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_path"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"/"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"tmp"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/c"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"rush"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/Test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Coder"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"Agent"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/open"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"router"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"imi"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"-k"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"2"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/write"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/write"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"_tool"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/config"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"/config"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".json"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":".json"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"content"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"content"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"{\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \"{\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"name"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"name"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"test"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\","},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"version"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"version"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\":"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":" \\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"1"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"1"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"0"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"0"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"."},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"0"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"0"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"\\\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"}\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"}\""},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":null,"tool_calls":[{"index":0,"function":{"arguments":"}"},"type":"function"}]},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"tool_calls","native_finish_reason":"tool_calls","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663468-CCuVCztqk1TuVZhnwWmf","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663469,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10939,"completion_tokens":53,"total_tokens":10992,"cost":0.0018575,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.001725,"upstream_inference_completions_cost":0.0001325},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669474-T5EEb1unXW3eyM4HyGAs","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669475,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10633,"completion_tokens":53,"total_tokens":10686,"cost":0.0017891,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0016566,"upstream_inference_completions_cost":0.0001325},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -172,15 +172,15 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 2.335058667s + duration: 2.596199041s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44776 + content_length: 43559 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"write:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool/config.json\\\", \\\"content\\\": \\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\"}\",\"name\":\"write\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\nFile successfully written: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool/config.json\\n\",\"tool_call_id\":\"write:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\",\"cache_control\":{\"type\":\"ephemeral\"}},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"tool_calls\":[{\"id\":\"write:0\",\"function\":{\"arguments\":\"{\\\"file_path\\\": \\\"/tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool/config.json\\\", \\\"content\\\": \\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\"}\",\"name\":\"write\"},\"type\":\"function\",\"cache_control\":{\"type\":\"ephemeral\"}}],\"role\":\"assistant\"},{\"content\":\"\\nFile successfully written: /tmp/crush-test/TestCoderAgent/openrouter-kimi-k2/write_tool/config.json\\n\",\"tool_call_id\":\"write:0\",\"role\":\"tool\"}],\"model\":\"moonshotai/kimi-k2-0905\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"usage\":{\"include\":true},\"stream\":true}" headers: Accept: - application/json @@ -196,13 +196,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"gen-1761663474-g4fX9062GSHSboswiX6j","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663474,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669480-AxGYjW9ua11XvzCE2laW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669480,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663474-g4fX9062GSHSboswiX6j","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663474,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669480-AxGYjW9ua11XvzCE2laW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669480,"choices":[{"index":0,"delta":{"role":"assistant","content":"Done"},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663474-g4fX9062GSHSboswiX6j","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663474,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} + data: {"id":"gen-1761669480-AxGYjW9ua11XvzCE2laW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669480,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":"stop","native_finish_reason":"stop","logprobs":null}],"system_fingerprint":"fpv0_41b1161c"} - data: {"id":"gen-1761663474-g4fX9062GSHSboswiX6j","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761663474,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":11037,"completion_tokens":2,"total_tokens":11039,"cost":0.0017888,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10752,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017838,"upstream_inference_completions_cost":0.000005},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} + data: {"id":"gen-1761669480-AxGYjW9ua11XvzCE2laW","provider":"Moonshot AI","model":"moonshotai/kimi-k2-0905","object":"chat.completion.chunk","created":1761669480,"choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null,"native_finish_reason":null,"logprobs":null}],"usage":{"prompt_tokens":10731,"completion_tokens":2,"total_tokens":10733,"cost":0.0017204,"is_byok":false,"prompt_tokens_details":{"cached_tokens":10496,"audio_tokens":0},"cost_details":{"upstream_inference_cost":null,"upstream_inference_prompt_cost":0.0017154,"upstream_inference_completions_cost":0.000005},"completion_tokens_details":{"reasoning_tokens":0,"image_tokens":0}}} data: [DONE] @@ -211,4 +211,4 @@ interactions: - text/event-stream status: 200 OK code: 200 - duration: 1.965889833s + duration: 1.788477167s diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/bash_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/bash_tool.yaml index 991553445b9f3b98386a14bd43a683d50d20f7cf..92ef56e30baa7f303a0a905b409b40538c0d484d 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/bash_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/bash_tool.yaml @@ -24,21 +24,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Create"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Create"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" bash"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" bash"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" hello"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" hello"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" content"}}]} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" content"}}]} - data: {"id":"202510282258553ebed55b95fd46eb","created":1761663535,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":134,"completion_tokens":10,"total_tokens":144,"prompt_tokens_details":{"cached_tokens":4}}} + data: {"id":"20251029003827459d360e992a4523","created":1761669507,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":134,"completion_tokens":10,"total_tokens":144,"prompt_tokens_details":{"cached_tokens":4}}} data: [DONE] @@ -47,15 +47,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 576.833ms + duration: 628.987834ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44149 + content_length: 42932 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -71,13 +71,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225855de584289cedb45ab","created":1761663535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"2025102900382734702fd46ed64597","created":1761669507,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225855de584289cedb45ab","created":1761663535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900382734702fd46ed64597","created":1761669507,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225855de584289cedb45ab","created":1761663535,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_fc8a5f5955c042fca263a86b","index":0,"type":"function","function":{"name":"bash","arguments":"{\"command\":\"echo 'hello bash' > test.txt\",\"description\":\"Create test.txt with content 'hello bash'\"}"}}]}}]} + data: {"id":"2025102900382734702fd46ed64597","created":1761669507,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_aba232121de7493098232b37","index":0,"type":"function","function":{"name":"bash","arguments":"{\"command\":\"echo 'hello bash' > test.txt\",\"description\":\"Create test.txt with content 'hello bash'\"}"}}]}}]} - data: {"id":"20251028225855de584289cedb45ab","created":1761663535,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10627,"completion_tokens":40,"total_tokens":10667,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"2025102900382734702fd46ed64597","created":1761669507,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10345,"completion_tokens":40,"total_tokens":10385,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -86,15 +86,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 660.371625ms + duration: 3.124087959s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44492 + content_length: 43275 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_fc8a5f5955c042fca263a86b\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"echo 'hello bash' > test.txt\\\",\\\"description\\\":\\\"Create test.txt with content 'hello bash'\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"call_fc8a5f5955c042fca263a86b\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_aba232121de7493098232b37\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"echo 'hello bash' > test.txt\\\",\\\"description\\\":\\\"Create test.txt with content 'hello bash'\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"call_aba232121de7493098232b37\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -110,13 +110,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225856c6dc9a7ce8464d38","created":1761663536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038308a8c2f8898404abc","created":1761669510,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225856c6dc9a7ce8464d38","created":1761663536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038308a8c2f8898404abc","created":1761669510,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225856c6dc9a7ce8464d38","created":1761663536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Done"}}]} + data: {"id":"202510290038308a8c2f8898404abc","created":1761669510,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_1851accb2d9a4de69eb49437","index":0,"type":"function","function":{"name":"bash","arguments":"{\"command\":\"cat test.txt\",\"description\":\"Verify file content\"}"}}]}}]} - data: {"id":"20251028225856c6dc9a7ce8464d38","created":1761663536,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10675,"completion_tokens":6,"total_tokens":10681,"prompt_tokens_details":{"cached_tokens":10666}}} + data: {"id":"202510290038308a8c2f8898404abc","created":1761669510,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10393,"completion_tokens":29,"total_tokens":10422,"prompt_tokens_details":{"cached_tokens":10384}}} data: [DONE] @@ -125,4 +125,61 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 710.481541ms + duration: 1.989138084s +- id: 3 + request: + proto: HTTP/1.1 + proto_major: 1 + proto_minor: 1 + content_length: 43649 + host: "" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/bash_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use bash to create a file named test.txt with content 'hello bash'\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_aba232121de7493098232b37\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"echo 'hello bash' > test.txt\\\",\\\"description\\\":\\\"Create test.txt with content 'hello bash'\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"no output\",\"tool_call_id\":\"call_aba232121de7493098232b37\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_1851accb2d9a4de69eb49437\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"cat test.txt\\\",\\\"description\\\":\\\"Verify file content\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"hello bash\\n\\n\\n/tmp/crush-test/TestCoderAgent/zai-glm4.6/bash_tool\",\"tool_call_id\":\"call_1851accb2d9a4de69eb49437\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + headers: + Accept: + - application/json + Content-Type: + - application/json + User-Agent: + - OpenAI/Go 2.7.1 + url: https://api.z.ai/api/coding/paas/v4/chat/completions + method: POST + response: + proto: HTTP/2.0 + proto_major: 2 + proto_minor: 0 + content_length: -1 + body: |+ + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Done"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" File"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" created"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" content"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" '"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"hello"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" bash"}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"'."}}]} + + data: {"id":"20251029003833dfc301949eda4b81","created":1761669513,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10452,"completion_tokens":15,"total_tokens":10467,"prompt_tokens_details":{"cached_tokens":45}}} + + data: [DONE] + + headers: + Content-Type: + - text/event-stream;charset=UTF-8 + status: 200 OK + code: 200 + duration: 4.192006292s diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/download_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/download_tool.yaml index 50ee185a16fed5ffc0e44288d2a7f1e3a02cdf11..d14f8d7df8a7f46ddebfb66110846ec3db11946a 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/download_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/download_tool.yaml @@ -6,9 +6,9 @@ interactions: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 778 + content_length: 42985 host: "" - body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\ndownload the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -24,21 +24,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Download"}}]} - - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" save"}}]} - - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" example"}}]} + data: {"id":"20251029003842b0b0628019ec41fc","created":1761669522,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":".txt"}}]} + data: {"id":"20251029003842b0b0628019ec41fc","created":1761669522,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} + data: {"id":"20251029003842b0b0628019ec41fc","created":1761669522,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_d5de08eaa52b4b93aa022e6b","index":0,"type":"function","function":{"name":"download","arguments":"{\"url\":\"https://example-files.online-convert.com/document/txt/example.txt\",\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool/example.txt\"}"}}]}}]} - data: {"id":"20251028225858b7808e8d25f7447c","created":1761663538,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":143,"completion_tokens":10,"total_tokens":153,"prompt_tokens_details":{"cached_tokens":4}}} + data: {"id":"20251029003842b0b0628019ec41fc","created":1761669522,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10353,"completion_tokens":55,"total_tokens":10408,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -47,15 +39,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 671.008458ms + duration: 581.707375ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44202 + content_length: 778 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\ndownload the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' headers: Accept: - application/json @@ -71,13 +63,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"2025102822585850e15d253c0e4109","created":1761663538,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Download"}}]} + + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" example"}}]} + + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"2025102822585850e15d253c0e4109","created":1761663538,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} - data: {"id":"2025102822585850e15d253c0e4109","created":1761663538,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_92e3b7db93a94e4db1179f6f","index":0,"type":"function","function":{"name":"download","arguments":"{\"url\":\"https://example-files.online-convert.com/document/txt/example.txt\",\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool/example.txt\"}"}}]}}]} + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" URL"}}]} - data: {"id":"2025102822585850e15d253c0e4109","created":1761663538,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10635,"completion_tokens":55,"total_tokens":10690,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"2025102900384241d8ae07e2b34a6d","created":1761669522,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":143,"completion_tokens":9,"total_tokens":152,"prompt_tokens_details":{"cached_tokens":22}}} data: [DONE] @@ -86,7 +84,7 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 783.232375ms + duration: 1.013031084s - id: 2 request: proto: HTTP/1.1 @@ -132,15 +130,15 @@ interactions: - text/plain; charset=UTF-8 status: 200 OK code: 200 - duration: 83.692417ms + duration: 154.745584ms - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44744 + content_length: 43527 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_92e3b7db93a94e4db1179f6f\",\"function\":{\"arguments\":\"{\\\"url\\\":\\\"https://example-files.online-convert.com/document/txt/example.txt\\\",\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool/example.txt\\\"}\",\"name\":\"download\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"call_92e3b7db93a94e4db1179f6f\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"download the file from https://example-files.online-convert.com/document/txt/example.txt and save it as example.txt\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_d5de08eaa52b4b93aa022e6b\",\"function\":{\"arguments\":\"{\\\"url\\\":\\\"https://example-files.online-convert.com/document/txt/example.txt\\\",\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool/example.txt\\\"}\",\"name\":\"download\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Successfully downloaded 2574 bytes to /tmp/crush-test/TestCoderAgent/zai-glm4.6/download_tool/example.txt (Content-Type: text/plain; charset=UTF-8)\",\"tool_call_id\":\"call_d5de08eaa52b4b93aa022e6b\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -156,25 +154,25 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Download"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"File"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"ed"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" downloaded"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" saved"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" saved"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" as"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" as"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" example"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" example"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".txt"}}]} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".txt"}}]} - data: {"id":"20251028225900fd23803d0ee242c1","created":1761663540,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10734,"completion_tokens":12,"total_tokens":10746,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"20251029003843a0f6b6232ba743eb","created":1761669523,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10452,"completion_tokens":12,"total_tokens":10464,"prompt_tokens_details":{"cached_tokens":45}}} data: [DONE] @@ -183,4 +181,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 714.331667ms + duration: 1.07914675s diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/glob_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/glob_tool.yaml index dd2f2b0f0b8a33dfaca9e4b066327da54a77b05b..9a2b18c38a498de4735c5dce6582684a87742450 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/glob_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/glob_tool.yaml @@ -24,21 +24,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Find"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Find"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" all"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" all"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" files"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" files"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" glob"}}]} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" glob"}}]} - data: {"id":"202510282259141882a6e5c66c4a52","created":1761663554,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":132,"completion_tokens":10,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":114}}} + data: {"id":"20251029003845fd2734afaac3455b","created":1761669525,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":132,"completion_tokens":10,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":114}}} data: [DONE] @@ -47,15 +47,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 935.680041ms + duration: 633.400792ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44138 + content_length: 42921 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -71,153 +71,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"2025102900384589512354d27741c8","created":1761669525,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"The"}}]} + data: {"id":"2025102900384589512354d27741c8","created":1761669525,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" user"}}]} + data: {"id":"2025102900384589512354d27741c8","created":1761669525,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_1b805df19e8f4d0e8d5af5c7","index":0,"type":"function","function":{"name":"glob","arguments":"{\"pattern\":\"*.go\"}"}}]}}]} - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" wants"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" use"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" glob"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tool"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" find"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" all"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" ."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"go"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" files"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" directory"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" This"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" is"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" straightforward"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" request"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" I"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" need"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":":\n\n"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"1"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Use"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" glob"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tool"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" with"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" pattern"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" \"*."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"go"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"\""}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" find"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" all"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Go"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" files"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"2"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" The"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" path"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" parameter"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" can"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" be"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" omitted"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" since"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" it"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" defaults"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" working"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" directory"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n\n"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Let"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" make"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tool"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" call"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_e48fa09a1f4e43f796145c00","index":0,"type":"function","function":{"name":"glob","arguments":"{\"pattern\":\"*.go\"}"}}]}}]} - - data: {"id":"20251028225914b1815711dc8e421b","created":1761663554,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10625,"completion_tokens":88,"total_tokens":10713,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"2025102900384589512354d27741c8","created":1761669525,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10343,"completion_tokens":18,"total_tokens":10361,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -226,15 +86,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 650.744167ms + duration: 862.482625ms - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44445 + content_length: 43228 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_e48fa09a1f4e43f796145c00\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/glob_tool/main.go\",\"tool_call_id\":\"call_e48fa09a1f4e43f796145c00\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/glob_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files in the current directory\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_1b805df19e8f4d0e8d5af5c7\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/glob_tool/main.go\",\"tool_call_id\":\"call_1b805df19e8f4d0e8d5af5c7\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -250,49 +110,27 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/tmp"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/c"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"rush"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"-test"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/Test"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Coder"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Agent"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/z"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"ai"}}]} - - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"-g"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"lm"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"4"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"6"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/g"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"lob"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"_tool"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/main"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"202510282259156d9acfc59a144376","created":1761663555,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10668,"completion_tokens":24,"total_tokens":10692,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290038463ae3c0ee81644c80","created":1761669526,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10386,"completion_tokens":13,"total_tokens":10399,"prompt_tokens_details":{"cached_tokens":10360}}} data: [DONE] @@ -301,4 +139,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 1.423900375s + duration: 585.200334ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/grep_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/grep_tool.yaml index ccf5469a9d3804c3f4839a11f559bc32afbcdaf2..c9a8468774a51a65b2442a4e035041742c09b43e 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/grep_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/grep_tool.yaml @@ -24,19 +24,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"grep"}}]} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"grep"}}]} - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" package"}}]} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" package"}}]} - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" go"}}]} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" go"}}]} - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" files"}}]} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" files"}}]} - data: {"id":"20251028225921ff8e208942d540d2","created":1761663561,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":133,"completion_tokens":9,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":4}}} + data: {"id":"2025102900384877bde624fb6a4946","created":1761669528,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":133,"completion_tokens":9,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":4}}} data: [DONE] @@ -45,15 +45,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 566.837833ms + duration: 867.963291ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44136 + content_length: 42919 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -69,13 +69,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225921617c057f76df41ef","created":1761663561,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"20251029003848037303600f934c86","created":1761669528,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225921617c057f76df41ef","created":1761663561,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003848037303600f934c86","created":1761669528,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225921617c057f76df41ef","created":1761663561,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_24ae82d2333b49b0ad0687c6","index":0,"type":"function","function":{"name":"grep","arguments":"{\"include\":\"*.go\",\"pattern\":\"package\"}"}}]}}]} + data: {"id":"20251029003848037303600f934c86","created":1761669528,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_cd2fb9b5c4284eb9a20e3d77","index":0,"type":"function","function":{"name":"grep","arguments":"{\"include\":\"*.go\",\"pattern\":\"package\"}"}}]}}]} - data: {"id":"20251028225921617c057f76df41ef","created":1761663561,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10626,"completion_tokens":26,"total_tokens":10652,"prompt_tokens_details":{"cached_tokens":10580}}} + data: {"id":"20251029003848037303600f934c86","created":1761669528,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10344,"completion_tokens":26,"total_tokens":10370,"prompt_tokens_details":{"cached_tokens":10298}}} data: [DONE] @@ -84,15 +84,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 667.12125ms + duration: 867.059375ms - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44519 + content_length: 43302 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_24ae82d2333b49b0ad0687c6\",\"function\":{\"arguments\":\"{\\\"include\\\":\\\"*.go\\\",\\\"pattern\\\":\\\"package\\\"}\",\"name\":\"grep\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/zai-glm4.6/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"call_24ae82d2333b49b0ad0687c6\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/grep_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use grep to search for the word 'package' in go files\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_cd2fb9b5c4284eb9a20e3d77\",\"function\":{\"arguments\":\"{\\\"include\\\":\\\"*.go\\\",\\\"pattern\\\":\\\"package\\\"}\",\"name\":\"grep\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Found 1 matches\\n/tmp/crush-test/TestCoderAgent/zai-glm4.6/grep_tool/main.go:\\n Line 1, Char 1: package main\\n\",\"tool_call_id\":\"call_cd2fb9b5c4284eb9a20e3d77\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -108,43 +108,77 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" match"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" match"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `/"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"main"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"tmp"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/c"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"rush"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"-test"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/Test"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" -"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Coder"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Agent"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"package"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/z"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"ai"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`"}}]} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"-g"}}]} - data: {"id":"20251028225922be006f9fca0e4676","created":1761663562,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10694,"completion_tokens":21,"total_tokens":10715,"prompt_tokens_details":{"cached_tokens":10651}}} + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"lm"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"4"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"6"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/g"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"rep"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"_tool"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/main"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" -"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \""}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"package"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\""}}]} + + data: {"id":"2025102900385045d0ac72ac7f41fe","created":1761669530,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10412,"completion_tokens":38,"total_tokens":10450,"prompt_tokens_details":{"cached_tokens":10369}}} data: [DONE] @@ -153,4 +187,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 1.377471625s + duration: 541.385ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/ls_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/ls_tool.yaml index adacd0361474f51e24ea7da58ad29bb7132ffdc1..efb83042eeeb53b890f147be779887b9850e24b1 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/ls_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/ls_tool.yaml @@ -6,9 +6,9 @@ interactions: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 712 + content_length: 42913 host: "" - body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nuse ls to list the files in the current directory\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -24,19 +24,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"List"}}]} - - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" files"}}]} - - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} + data: {"id":"202510290038526a99a791ab294e56","created":1761669532,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" ls"}}]} + data: {"id":"202510290038526a99a791ab294e56","created":1761669532,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" command"}}]} + data: {"id":"202510290038526a99a791ab294e56","created":1761669532,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_bf3de6604dda426da72dd2c6","index":0,"type":"function","function":{"name":"ls","arguments":"{\"path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\"}"}}]}}]} - data: {"id":"202510282259249c24d6a539d34fa4","created":1761663564,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":130,"completion_tokens":9,"total_tokens":139,"prompt_tokens_details":{"cached_tokens":115}}} + data: {"id":"202510290038526a99a791ab294e56","created":1761669532,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10341,"completion_tokens":33,"total_tokens":10374,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -45,15 +39,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 746.722333ms + duration: 570.890125ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44130 + content_length: 712 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nuse ls to list the files in the current directory\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' headers: Accept: - application/json @@ -69,13 +63,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259250a666966d1d54d83","created":1761663565,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"List"}}]} + + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" files"}}]} + + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"202510282259250a666966d1d54d83","created":1761663565,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" ls"}}]} - data: {"id":"202510282259250a666966d1d54d83","created":1761663565,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_c69851fefa134668ba3611c1","index":0,"type":"function","function":{"name":"ls","arguments":"{\"path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\"}"}}]}}]} + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" command"}}]} - data: {"id":"202510282259250a666966d1d54d83","created":1761663565,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10623,"completion_tokens":33,"total_tokens":10656,"prompt_tokens_details":{"cached_tokens":43}}} + data: {"id":"20251029003852c0d24590439d4402","created":1761669532,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":130,"completion_tokens":9,"total_tokens":139,"prompt_tokens_details":{"cached_tokens":4}}} data: [DONE] @@ -84,15 +84,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 2.1793745s + duration: 760.08475ms - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44499 + content_length: 43282 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_c69851fefa134668ba3611c1\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_c69851fefa134668ba3611c1\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use ls to list the files in the current directory\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_bf3de6604dda426da72dd2c6\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/zai-glm4.6/ls_tool/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_bf3de6604dda426da72dd2c6\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -108,21 +108,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".mod"}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".mod"}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"202510282259276c55fb550f8d41f7","created":1761663567,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10691,"completion_tokens":10,"total_tokens":10701,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"2025102900385396a006eadd2e483a","created":1761669533,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10409,"completion_tokens":10,"total_tokens":10419,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -131,4 +131,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 1.01095175s + duration: 574.576667ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/multiedit_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/multiedit_tool.yaml index f24af44d33dcc9b43fbce7986147297f984b9b85..a847870ba2ef3a77a32e8bb3681c262e9252c1ed 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/multiedit_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/multiedit_tool.yaml @@ -24,21 +24,23 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Mult"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Edit"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"ied"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"it"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" code"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" code"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" modification"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" mult"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" request"}}]} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"ied"}}]} - data: {"id":"20251028225928a02e3f15bfe14c6a","created":1761663568,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":153,"completion_tokens":10,"total_tokens":163,"prompt_tokens_details":{"cached_tokens":115}}} + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"it"}}]} + + data: {"id":"2025102900385425dddcc8ab834e11","created":1761669534,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":153,"completion_tokens":11,"total_tokens":164,"prompt_tokens_details":{"cached_tokens":4}}} data: [DONE] @@ -47,15 +49,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 685.060333ms + duration: 508.653ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44216 + content_length: 42999 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -71,149 +73,174 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"The"}}]} - - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" user"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" wants"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"I"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"'ll"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" use"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" help"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" you"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" mult"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" use"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"ied"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" mult"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"it"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"ied"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tool"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"it"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":":\n"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" modify"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"1"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" the"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Change"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" '"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Hello"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":","}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Let"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" World"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" me"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"!'"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" first"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" read"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" '"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" the"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Hello"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":","}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Crush"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" understand"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"!'\n"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" its"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"2"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" current"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" structure"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Add"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".\n"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_b7403cb7152f492692af49d1","index":0,"type":"function","function":{"name":"view","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\"}"}}]}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" comment"}}]} + data: {"id":"2025102900385482acb57e53794a18","created":1761669535,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10366,"completion_tokens":65,"total_tokens":10431,"prompt_tokens_details":{"cached_tokens":10297}}} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" '//"}}]} + data: [DONE] - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" G"}}]} + headers: + Content-Type: + - text/event-stream;charset=UTF-8 + status: 200 OK + code: 200 + duration: 575.737833ms +- id: 2 + request: + proto: HTTP/1.1 + proto_major: 1 + proto_minor: 1 + content_length: 43582 + host: "" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"\\nI'll help you use multiedit to modify the main.go file. Let me first read the file to understand its current structure.\\n\",\"tool_calls\":[{\"id\":\"call_b7403cb7152f492692af49d1\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_b7403cb7152f492692af49d1\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + headers: + Accept: + - application/json + Content-Type: + - application/json + User-Agent: + - OpenAI/Go 2.7.1 + url: https://api.z.ai/api/coding/paas/v4/chat/completions + method: POST + response: + proto: HTTP/2.0 + proto_major: 2 + proto_minor: 0 + content_length: -1 + body: |+ + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"reeting"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"'"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Now"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" above"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" I"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"'ll"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" fmt"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" use"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".Println"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" mult"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" line"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"ied"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"it"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" make"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n\n"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" both"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"First"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" changes"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":","}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" I"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" replace"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" need"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \""}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" read"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" World"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!\""}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \""}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" see"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Crush"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!\""}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" content"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" and"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" add"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" understand"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" the"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" comment"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" exact"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" above"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" formatting"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" the"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" and"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" fmt"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" indentation"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".Println"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" line"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".\n"}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_a6915eda77d14eb1a2a02982","index":0,"type":"function","function":{"name":"view","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\"}"}}]}}]} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_84768b5d83754b77b0089ceb","index":0,"type":"function","function":{"name":"multiedit","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\",\"edits\":[{\"old_string\": \"func main() {\\n\\tfmt.Println(\\\"Hello, World!\\\")\\n}\", \"new_string\": \"func main() {\\n\\t// Greeting\\n\\tfmt.Println(\\\"Hello, Crush!\\\")\\n}\"}]}"}}]}}]} - data: {"id":"20251028225928f800abfa305d4d5d","created":1761663568,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10648,"completion_tokens":106,"total_tokens":10754,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"20251029003856e4cdb8879ba6499c","created":1761669536,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10488,"completion_tokens":134,"total_tokens":10622,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -222,15 +249,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 630.192333ms -- id: 2 + duration: 1.489305s +- id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44678 + content_length: 44340 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_a6915eda77d14eb1a2a02982\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_a6915eda77d14eb1a2a02982\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"\\nI'll help you use multiedit to modify the main.go file. Let me first read the file to understand its current structure.\\n\",\"tool_calls\":[{\"id\":\"call_b7403cb7152f492692af49d1\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_b7403cb7152f492692af49d1\",\"role\":\"tool\"},{\"content\":\"\\nNow I'll use multiedit to make both changes: replace \\\"Hello, World!\\\" with \\\"Hello, Crush!\\\" and add the comment above the fmt.Println line.\\n\",\"tool_calls\":[{\"id\":\"call_84768b5d83754b77b0089ceb\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\\",\\\"edits\\\":[{\\\"old_string\\\": \\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\", \\\"new_string\\\": \\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\"}]}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"call_84768b5d83754b77b0089ceb\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -246,108 +273,93 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259302329ae1387054692","created":1761663570,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282259302329ae1387054692","created":1761663570,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259302329ae1387054692","created":1761663570,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_8695f45b2ab845c199121578","index":0,"type":"function","function":{"name":"multiedit","arguments":"{\"edits\":[{\"new_string\": \"func main() {\\n\\t// Greeting\\n\\tfmt.Println(\\\"Hello, Crush!\\\")\\n}\", \"old_string\": \"func main() {\\n\\tfmt.Println(\\\"Hello, World!\\\")\\n}\"}],\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\"}"}}]}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Done"}}]} - data: {"id":"202510282259302329ae1387054692","created":1761663570,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10743,"completion_tokens":100,"total_tokens":10843,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!"}}]} - data: [DONE] + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" I"}}]} - headers: - Content-Type: - - text/event-stream;charset=UTF-8 - status: 200 OK - code: 200 - duration: 827.610458ms -- id: 3 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 45293 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use multiedit to change 'Hello, World!' to 'Hello, Crush!' and add a comment '// Greeting' above the fmt.Println line in main.go\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_a6915eda77d14eb1a2a02982\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_a6915eda77d14eb1a2a02982\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_8695f45b2ab845c199121578\",\"function\":{\"arguments\":\"{\\\"edits\\\":[{\\\"new_string\\\": \\\"func main() {\\\\n\\\\t// Greeting\\\\n\\\\tfmt.Println(\\\\\\\"Hello, Crush!\\\\\\\")\\\\n}\\\", \\\"old_string\\\": \\\"func main() {\\\\n\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\\n}\\\"}],\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\\"}\",\"name\":\"multiedit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nApplied 1 edits to file: /tmp/crush-test/TestCoderAgent/zai-glm4.6/multiedit_tool/main.go\\n\\n\",\"tool_call_id\":\"call_8695f45b2ab845c199121578\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://api.z.ai/api/coding/paas/v4/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"'ve"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" successfully"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" used"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" mult"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"ied"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"it"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} + + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" change"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \""}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Done"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Changed"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" World"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" '"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!\""}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \""}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" World"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!'"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Crush"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" '"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!\""}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" added"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Crush"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" the"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!'"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \"//"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" G"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" added"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"reeting"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" '//"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\""}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" G"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" comment"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"reeting"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" above"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"'"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" the"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" comment"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" fmt"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".Println"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" line"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"5"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"-"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"6"}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"6"}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} - data: {"id":"2025102822593144b780ac4ea448bf","created":1761663571,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10884,"completion_tokens":34,"total_tokens":10918,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290038586b9996ac7e2c458a","created":1761669539,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10663,"completion_tokens":46,"total_tokens":10709,"prompt_tokens_details":{"cached_tokens":10430}}} data: [DONE] @@ -356,4 +368,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 10.714352666s + duration: 866.237083ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/parallel_tool_calls.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/parallel_tool_calls.yaml index 275934123b027e27d7c1505249c68be9ba60602f..95e4abdbe7656c5480001c65bf304f7464e6f8e3 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/parallel_tool_calls.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/parallel_tool_calls.yaml @@ -6,9 +6,9 @@ interactions: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 797 + content_length: 43010 host: "" - body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nuse glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -24,19 +24,15 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Parallel"}}]} + data: {"id":"20251029003909df7e852c93f54737","created":1761669549,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" glob"}}]} + data: {"id":"20251029003909df7e852c93f54737","created":1761669549,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} + data: {"id":"20251029003909df7e852c93f54737","created":1761669549,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_0f0e89a4bd5e41559a721f92","index":0,"type":"function","function":{"name":"glob","arguments":"{\"pattern\":\"**/*.go\"}"}}]}}]} - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" ls"}}]} + data: {"id":"20251029003909df7e852c93f54737","created":1761669549,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_b5efddc2502146e5bc8e7a22","index":1,"type":"function","function":{"name":"ls","arguments":"{\"path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\"}"}}]}}]} - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" commands"}}]} - - data: {"id":"202510282259507b4c984ee9aa4ff2","created":1761663590,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":149,"completion_tokens":9,"total_tokens":158,"prompt_tokens_details":{"cached_tokens":122}}} + data: {"id":"20251029003909df7e852c93f54737","created":1761669549,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10361,"completion_tokens":49,"total_tokens":10410,"prompt_tokens_details":{"cached_tokens":10298}}} data: [DONE] @@ -45,15 +41,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 667.117209ms + duration: 612.312958ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44227 + content_length: 797 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nuse glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' headers: Accept: - application/json @@ -69,133 +65,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"The"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" user"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" wants"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":":\n"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"1"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Use"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" glob"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" find"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" all"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" ."}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"go"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" files"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"2"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Use"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" ls"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" list"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" directory"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"3"}}]} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Parallel"}}]} - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Run"}}]} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" glob"}}]} - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" both"}}]} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tool"}}]} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" ls"}}]} - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" calls"}}]} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" commands"}}]} - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" parallel"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n\n"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"I"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" need"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" use"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" both"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tools"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" single"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" message"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" run"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" them"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" parallel"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Let"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" make"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" tool"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" calls"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_90c4d294870442d49800c870","index":0,"type":"function","function":{"name":"glob","arguments":"{\"pattern\":\"**/*.go\"}"}}]}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_95f4a5a97f3c4b0987e5e6be","index":1,"type":"function","function":{"name":"ls","arguments":"{\"path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\"}"}}]}}]} - - data: {"id":"20251028225950ce1e99bff28a4df4","created":1761663590,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10643,"completion_tokens":108,"total_tokens":10751,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"2025102900390946370129114d4422","created":1761669549,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":149,"completion_tokens":9,"total_tokens":158,"prompt_tokens_details":{"cached_tokens":115}}} data: [DONE] @@ -204,15 +86,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 604.815916ms + duration: 740.142083ms - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44889 + content_length: 43672 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_90c4d294870442d49800c870\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"call_95f4a5a97f3c4b0987e5e6be\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls/main.go\",\"tool_call_id\":\"call_90c4d294870442d49800c870\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_95f4a5a97f3c4b0987e5e6be\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use glob to find all .go files and use ls to list the current directory, it is very important that you run both tool calls in parallel\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_0f0e89a4bd5e41559a721f92\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/*.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"},{\"id\":\"call_b5efddc2502146e5bc8e7a22\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls/main.go\",\"tool_call_id\":\"call_0f0e89a4bd5e41559a721f92\",\"role\":\"tool\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/zai-glm4.6/parallel_tool_calls/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_b5efddc2502146e5bc8e7a22\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -228,47 +110,55 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} + + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} + + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" ."}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" ."}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"main"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":":"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`."}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Directory"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" contains"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Current"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" directory"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".mod"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" contains"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" go"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".mod"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"main"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`."}}]} - data: {"id":"20251028225952e294f5b79e4849c6","created":1761663592,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10753,"completion_tokens":23,"total_tokens":10776,"prompt_tokens_details":{"cached_tokens":10645}}} + data: {"id":"20251029003910d433885a9e1f4e09","created":1761669550,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10471,"completion_tokens":27,"total_tokens":10498,"prompt_tokens_details":{"cached_tokens":45}}} data: [DONE] @@ -277,4 +167,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 690.978584ms + duration: 1.155345375s diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/read_a_file.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/read_a_file.yaml index 69782544502a49926b4533194fab299a940bcf3f..393e11a00f215748164a0e67d610b03bd288a38a 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/read_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/read_a_file.yaml @@ -6,9 +6,9 @@ interactions: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 678 + content_length: 42883 host: "" - body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nRead the go mod\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -24,15 +24,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225806518c6376c29b42d8","created":1761663487,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003819d8412bfbd067436e","created":1761669499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225806518c6376c29b42d8","created":1761663487,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Go"}}]} + data: {"id":"20251029003819d8412bfbd067436e","created":1761669499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225806518c6376c29b42d8","created":1761663487,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" mod"}}]} + data: {"id":"20251029003819d8412bfbd067436e","created":1761669499,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_3c5bf30df2b9471a925e72af","index":0,"type":"function","function":{"name":"view","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file/go.mod\"}"}}]}}]} - data: {"id":"20251028225806518c6376c29b42d8","created":1761663487,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" read"}}]} - - data: {"id":"20251028225806518c6376c29b42d8","created":1761663487,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":124,"completion_tokens":7,"total_tokens":131,"prompt_tokens_details":{"cached_tokens":114}}} + data: {"id":"20251029003819d8412bfbd067436e","created":1761669499,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10335,"completion_tokens":36,"total_tokens":10371,"prompt_tokens_details":{"cached_tokens":45}}} data: [DONE] @@ -41,15 +39,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 563.050625ms + duration: 1.217761125s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44100 + content_length: 678 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nRead the go mod\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' headers: Accept: - application/json @@ -65,155 +63,15 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"The"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" user"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" wants"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" read"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" go"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" mod"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" This"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" is"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" typically"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" `"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"go"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".mod"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"`"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Go"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" project"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" I"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" need"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" first"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" check"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" what"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" files"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" are"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" available"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" directory"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" find"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" go"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".mod"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":","}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" then"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" read"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" its"}}]} + data: {"id":"20251029003819f65c3a300d3247f1","created":1761669499,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" contents"}}]} + data: {"id":"20251029003819f65c3a300d3247f1","created":1761669499,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Read"}}]} - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".\n\n"}}]} + data: {"id":"20251029003819f65c3a300d3247f1","created":1761669499,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" go"}}]} - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Let"}}]} + data: {"id":"20251029003819f65c3a300d3247f1","created":1761669499,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" mod"}}]} - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" start"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" by"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" listing"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" files"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" in"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" directory"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" see"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" if"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" there"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"'s"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" go"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".mod"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_d2b1319d790e410b8fd7ece6","index":0,"type":"function","function":{"name":"ls","arguments":"{\"path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\"}"}}]}}]} - - data: {"id":"202510282258074fca18a4bc684551","created":1761663487,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10617,"completion_tokens":104,"total_tokens":10721,"prompt_tokens_details":{"cached_tokens":45}}} + data: {"id":"20251029003819f65c3a300d3247f1","created":1761669499,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":124,"completion_tokens":7,"total_tokens":131,"prompt_tokens_details":{"cached_tokens":114}}} data: [DONE] @@ -222,54 +80,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 3.636547083s + duration: 1.305073541s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44477 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_d2b1319d790e410b8fd7ece6\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_d2b1319d790e410b8fd7ece6\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://api.z.ai/api/coding/paas/v4/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"20251028225814b3fc6d6a7d71410b","created":1761663495,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225814b3fc6d6a7d71410b","created":1761663495,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225814b3fc6d6a7d71410b","created":1761663495,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_abc89496fbf4431aa7a18349","index":0,"type":"function","function":{"name":"view","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file/go.mod\"}"}}]}}]} - - data: {"id":"20251028225814b3fc6d6a7d71410b","created":1761663495,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10685,"completion_tokens":36,"total_tokens":10721,"prompt_tokens_details":{"cached_tokens":10579}}} - - data: [DONE] - - headers: - Content-Type: - - text/event-stream;charset=UTF-8 - status: 200 OK - code: 200 - duration: 3.977741917s -- id: 3 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 44864 + content_length: 43270 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_d2b1319d790e410b8fd7ece6\",\"function\":{\"arguments\":\"{\\\"path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\\"}\",\"name\":\"ls\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n- /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file/\\n - go.mod\\n - main.go\\n\",\"tool_call_id\":\"call_d2b1319d790e410b8fd7ece6\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_abc89496fbf4431aa7a18349\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\n\",\"tool_call_id\":\"call_abc89496fbf4431aa7a18349\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Read the go mod\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_3c5bf30df2b9471a925e72af\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/read_a_file/go.mod\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|module example.com/testproject\\n 2|\\n 3|go 1.23\\n\\n\",\"tool_call_id\":\"call_3c5bf30df2b9471a925e72af\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -285,33 +104,33 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"module"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"module"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" example"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" example"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".com"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".com"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/test"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"/test"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"project"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"project"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"go"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"1"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"23"}}]} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"23"}}]} - data: {"id":"2025102822581992bf77a303e04c61","created":1761663499,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10756,"completion_tokens":16,"total_tokens":10772,"prompt_tokens_details":{"cached_tokens":10720}}} + data: {"id":"20251029003821314a74188014444b","created":1761669501,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10406,"completion_tokens":16,"total_tokens":10422,"prompt_tokens_details":{"cached_tokens":10370}}} data: [DONE] @@ -320,4 +139,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 682.688792ms + duration: 801.871333ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/simple_test.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/simple_test.yaml index 3fd0d234edf1c01a4168d2b99f739f6761700668..c58f29500e91c708ac2032bd24c6a5730b9e203a 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/simple_test.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/simple_test.yaml @@ -24,13 +24,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225804960ed04de2a34ff6","created":1761663484,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003810cbf9425ae57b43ed","created":1761669490,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225804960ed04de2a34ff6","created":1761663484,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Simple"}}]} + data: {"id":"20251029003810cbf9425ae57b43ed","created":1761669490,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"G"}}]} - data: {"id":"20251028225804960ed04de2a34ff6","created":1761663484,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" greeting"}}]} + data: {"id":"20251029003810cbf9425ae57b43ed","created":1761669490,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"reeting"}}]} - data: {"id":"20251028225804960ed04de2a34ff6","created":1761663484,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":121,"completion_tokens":6,"total_tokens":127,"prompt_tokens_details":{"cached_tokens":4}}} + data: {"id":"20251029003810cbf9425ae57b43ed","created":1761669490,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":121,"completion_tokens":6,"total_tokens":127,"prompt_tokens_details":{"cached_tokens":114}}} data: [DONE] @@ -39,15 +39,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 1.431727375s + duration: 1.327415791s - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44090 + content_length: 42873 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Hello\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/simple_test\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"Hello\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -63,29 +63,53 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"!"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" How"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" I"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" can"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"'m"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" I"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" here"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" help"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" you"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" help"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" today"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"?"}}]} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" your"}}]} - data: {"id":"2025102822580569bb1503642e4d8b","created":1761663485,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10613,"completion_tokens":14,"total_tokens":10627,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" programming"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" tasks"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" What"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" would"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" you"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" like"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" me"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" work"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" on"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"?"}}]} + + data: {"id":"202510290038106d6cb6f172784153","created":1761669490,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10331,"completion_tokens":26,"total_tokens":10357,"prompt_tokens_details":{"cached_tokens":45}}} data: [DONE] @@ -94,4 +118,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 1.941200208s + duration: 8.493175125s diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/sourcegraph_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/sourcegraph_tool.yaml index 510b3c9da9f7cf4367cacf3fb9fec8b154c770a0..69240cf43844fcff0e5703122e09661a8fc24af1 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/sourcegraph_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/sourcegraph_tool.yaml @@ -6,9 +6,9 @@ interactions: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 723 + content_length: 42933 host: "" - body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nuse sourcegraph to search for ''func main'' in Go repositories\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -24,19 +24,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"search"}}]} + data: {"id":"2025102900390077992afa4d2144d0","created":1761669540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" func"}}]} + data: {"id":"2025102900390077992afa4d2144d0","created":1761669540,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"2025102900390077992afa4d2144d0","created":1761669540,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_46968e1bd0c348ef833889b6","index":0,"type":"function","function":{"name":"sourcegraph","arguments":"{\"query\":\"lang:go func main\",\"count\":10}"}}]}}]} - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} - - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} - - data: {"id":"20251028225943469524737f404ad1","created":1761663583,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":133,"completion_tokens":9,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":115}}} + data: {"id":"2025102900390077992afa4d2144d0","created":1761669540,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10344,"completion_tokens":30,"total_tokens":10374,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -45,15 +39,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 634.14525ms + duration: 696.6315ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44150 + content_length: 723 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: '{"messages":[{"content":"you will generate a short title based on the first message a user begins a conversation with\n\n\n- ensure it is not more than 50 characters long\n- the title should be a summary of the user''s message\n- it should be one line long\n- do not use quotes or colons\n- the entire text you return will be used as the title\n- never return anything that is more than one sentence (one line) long\n\n\n /no_think","role":"system"},{"content":"Generate a concise title for the following content:\n\nuse sourcegraph to search for ''func main'' in Go repositories\n \n\n","role":"user"}],"model":"glm-4.5-air","max_tokens":40,"stream_options":{"include_usage":true},"stream":true}' headers: Accept: - application/json @@ -69,13 +63,19 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225943603311c911f04425","created":1761663583,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Searching"}}]} + + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" func"}}]} - data: {"id":"20251028225943603311c911f04425","created":1761663583,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"20251028225943603311c911f04425","created":1761663583,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_2373c4cfefc140738398b8e6","index":0,"type":"function","function":{"name":"sourcegraph","arguments":"{\"query\":\"lang:go func main\"}"}}]}}]} + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} - data: {"id":"20251028225943603311c911f04425","created":1761663583,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10626,"completion_tokens":22,"total_tokens":10648,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} + + data: {"id":"2025102900390068c384ebb8f44f73","created":1761669540,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":133,"completion_tokens":9,"total_tokens":142,"prompt_tokens_details":{"cached_tokens":115}}} data: [DONE] @@ -84,7 +84,7 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 630.015208ms + duration: 840.110708ms - id: 2 request: proto: HTTP/1.1 @@ -112,15 +112,15 @@ interactions: - application/json status: 200 OK code: 200 - duration: 1.104417125s + duration: 1.257034834s - id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 51878 + content_length: 50674 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_2373c4cfefc140738398b8e6\",\"function\":{\"arguments\":\"{\\\"query\\\":\\\"lang:go func main\\\"}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| func main() {\\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func greeting() {\\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n## Result 2: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/02_func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 3: github.com/jcalabro/uscope/assets/gobacktrace/main.go\\n\\nURL: /github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go\\n\\n```\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| func main() {\\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| \\tFuncA()\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncB()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncC()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncD()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncE()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncF()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func FuncE() {\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\tfmt.Println(\\\"FuncE\\\")\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| func FuncD() {\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| \\tFuncE()\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tfmt.Println(\\\"FuncD\\\")\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n```\\n\\n```\\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| func FuncC() {\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n```\\n\\n```\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| \\tFuncD()\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n```\\n\\n```\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tfmt.Println(\\\"FuncC\\\")\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n```\\n\\n```\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| func FuncB() {\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n```\\n\\n```\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| \\tFuncC()\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n```\\n\\n```\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tfmt.Println(\\\"FuncB\\\")\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n```\\n\\n```\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| func FuncA() {\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n```\\n\\n\",\"tool_call_id\":\"call_2373c4cfefc140738398b8e6\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/sourcegraph_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use sourcegraph to search for 'func main' in Go repositories\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_46968e1bd0c348ef833889b6\",\"function\":{\"arguments\":\"{\\\"query\\\":\\\"lang:go func main\\\",\\\"count\\\":10}\",\"name\":\"sourcegraph\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"# Sourcegraph Search Results\\n\\nFound 30 matches across 30 results\\n(Result limit reached, try a more specific query)\\n\\n## Result 1: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/01_before-func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| func main() {\\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func greeting() {\\n5| func greeting() {\\n6| \\tfmt.Println(\\\"Hello world!\\\")\\n7| }\\n8| \\n9| func main() {\\n10| \\tgreeting()\\n11| }\\n12| \\n```\\n\\n## Result 2: github.com/GoesToEleven/GolangTraining/14_functions/10_func-expression/02_func-expression/main.go\\n\\nURL: /github.com/GoesToEleven/GolangTraining/-/blob/14_functions/10_func-expression/02_func-expression/main.go\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func main() {\\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\tgreeting := func() {\\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n11| \\tgreeting()\\n12| }\\n13| \\n```\\n\\n```\\n0| package main\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func main() {\\n6| \\n7| \\tgreeting := func() {\\n8| \\t\\tfmt.Println(\\\"Hello world!\\\")\\n9| \\t}\\n10| \\n```\\n\\n## Result 3: github.com/jcalabro/uscope/assets/gobacktrace/main.go\\n\\nURL: /github.com/jcalabro/uscope/-/blob/assets/gobacktrace/main.go\\n\\n```\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| func main() {\\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| \\tFuncA()\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncB()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncC()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncD()\\n38| \\tFuncD()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncE()\\n39| \\tFuncE()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n34| func main() {\\n35| \\tFuncA()\\n36| \\tFuncB()\\n37| \\tFuncC()\\n38| \\tFuncD()\\n39| \\tFuncF()\\n40| \\tFuncF()\\n41| }\\n42| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| func FuncE() {\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| \\tfmt.Println(\\\"FuncE\\\")\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| func FuncD() {\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| \\tFuncE()\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n```\\n\\n```\\n1| package main\\n2| \\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tfmt.Println(\\\"FuncD\\\")\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n```\\n\\n```\\n3| import \\\"fmt\\\"\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| func FuncC() {\\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n```\\n\\n```\\n4| \\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| \\tFuncD()\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n```\\n\\n```\\n5| func FuncE() {\\n6| \\tfmt.Println(\\\"FuncE\\\")\\n7| }\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tfmt.Println(\\\"FuncC\\\")\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n```\\n\\n```\\n8| \\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| func FuncB() {\\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n```\\n\\n```\\n9| func FuncD() {\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| \\tFuncC()\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n```\\n\\n```\\n10| \\tFuncE()\\n11| \\tfmt.Println(\\\"FuncD\\\")\\n12| }\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tfmt.Println(\\\"FuncB\\\")\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| \\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n```\\n\\n```\\n13| \\n14| func FuncC() {\\n15| \\tFuncD()\\n16| \\tfmt.Println(\\\"FuncC\\\")\\n17| }\\n18| \\n19| func FuncB() {\\n20| \\tFuncC()\\n21| \\tfmt.Println(\\\"FuncB\\\")\\n22| }\\n23| func FuncA() {\\n24| func FuncA() {\\n25| \\tFuncB()\\n26| \\tfmt.Println(\\\"FuncA\\\")\\n27| }\\n28| \\n29| func FuncF() {\\n30| \\tfmt.Println(\\\"FuncF\\\")\\n31| \\tFuncE()\\n32| }\\n33| \\n```\\n\\n\",\"tool_call_id\":\"call_46968e1bd0c348ef833889b6\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -136,75 +136,103 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"30"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" matches"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" across"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Results"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" include"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" functions"}}]} + + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" training"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Found"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" "}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"30"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" testing"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" matches"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" tools"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" for"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" `"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"func"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" various"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"`"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" applications"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" in"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" The"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" search"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" shows"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Results"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" different"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" include"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" patterns"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" training"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" of"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" examples"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" function"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" various"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" implementations"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Go"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":","}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" programs"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" simple"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" repositories"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" hello"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" like"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" world"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" G"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" examples"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"olang"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Training"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" more"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" complex"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" us"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" function"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"cope"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" calling"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" projects"}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" patterns"}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} - data: {"id":"202510282259455a7f31f6052d4be7","created":1761663585,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":13356,"completion_tokens":37,"total_tokens":13393,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290039025231ff4296db4d2a","created":1761669542,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":13082,"completion_tokens":51,"total_tokens":13133,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -213,4 +241,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 919.849042ms + duration: 813.438042ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/update_a_file.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/update_a_file.yaml index 410780eb460b233c544a0686afeba8084334b773..b811e190e098b39316b960e5854daf7e21602a11 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/update_a_file.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/update_a_file.yaml @@ -24,25 +24,23 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Update"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Update"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" print"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" hello"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" hello"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" crush"}}]} - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" crush"}}]} - - data: {"id":"2025102822582113fc68da71c44ae7","created":1761663501,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":134,"completion_tokens":12,"total_tokens":146,"prompt_tokens_details":{"cached_tokens":114}}} + data: {"id":"20251029003822bd33d51dc1904eef","created":1761669502,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":134,"completion_tokens":11,"total_tokens":145,"prompt_tokens_details":{"cached_tokens":4}}} data: [DONE] @@ -51,15 +49,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 916.907458ms + duration: 774.391083ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44156 + content_length: 42939 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -75,255 +73,189 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"The"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"The"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" user"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" user"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" wants"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" wants"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" update"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" update"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" by"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" by"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" changing"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" changing"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" a"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" print"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" print"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" statement"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" statement"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" say"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" say"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" \""}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" \""}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"hello"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"hello"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" from"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" from"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" crush"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" crush"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"\"."}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"\"."}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" I"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" I"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" need"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" need"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":":\n\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":":\n\n"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"1"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"1"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" First"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" First"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":","}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" read"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" find"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" see"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"2"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" its"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Read"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" contents"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" it"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"2"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" see"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" what"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Find"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" current"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" print"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" print"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" statement"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" statement"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" that"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" looks"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" needs"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" like"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" be"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"3"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" changed"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Edit"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"3"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" it"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Edit"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" change"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" it"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" print"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" say"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" \""}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" \""}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"hello"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"hello"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" from"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" from"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" crush"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" crush"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"\"\n"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"\"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"4"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"4"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Verify"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" Test"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" change"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" change"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" was"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" if"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" made"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" possible"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" correctly"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n\n"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Let"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"Let"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" me"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" start"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" start"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" by"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" by"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" finding"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" reading"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" the"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" main"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":".go"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" file"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" to"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" see"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_1dbca8706f7c4525b02a0169","index":0,"type":"function","function":{"name":"glob","arguments":"{\"pattern\":\"**/main.go\"}"}}]}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" what"}}]} - data: {"id":"202510282258217f7db04cef4c47dd","created":1761663501,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10627,"completion_tokens":102,"total_tokens":10729,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"'s"}}]} - data: [DONE] + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" currently"}}]} - headers: - Content-Type: - - text/event-stream;charset=UTF-8 - status: 200 OK - code: 200 - duration: 15.323525958s -- id: 2 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 44473 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_1dbca8706f7c4525b02a0169\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\",\"tool_call_id\":\"call_1dbca8706f7c4525b02a0169\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://api.z.ai/api/coding/paas/v4/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"20251028225848202f414c561b4ba6","created":1761663528,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225848202f414c561b4ba6","created":1761663528,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225848202f414c561b4ba6","created":1761663528,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_17ae9185694a4283b7a3ae03","index":0,"type":"function","function":{"name":"view","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\"}"}}]}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":" there"}}]} - data: {"id":"20251028225848202f414c561b4ba6","created":1761663528,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10672,"completion_tokens":36,"total_tokens":10708,"prompt_tokens_details":{"cached_tokens":10579}}} - - data: [DONE] - - headers: - Content-Type: - - text/event-stream;charset=UTF-8 - status: 200 OK - code: 200 - duration: 632.079875ms -- id: 3 - request: - proto: HTTP/1.1 - proto_major: 1 - proto_minor: 1 - content_length: 44934 - host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_1dbca8706f7c4525b02a0169\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\",\"tool_call_id\":\"call_1dbca8706f7c4525b02a0169\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_17ae9185694a4283b7a3ae03\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_17ae9185694a4283b7a3ae03\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" - headers: - Accept: - - application/json - Content-Type: - - application/json - User-Agent: - - OpenAI/Go 2.7.1 - url: https://api.z.ai/api/coding/paas/v4/chat/completions - method: POST - response: - proto: HTTP/2.0 - proto_major: 2 - proto_minor: 0 - content_length: -1 - body: |+ - data: {"id":"20251028225849bbce80c093624e4b","created":1761663529,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","reasoning_content":"."}}]} - data: {"id":"20251028225849bbce80c093624e4b","created":1761663529,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225849bbce80c093624e4b","created":1761663529,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_3d7f707f9a9d49a19e405fa9","index":0,"type":"function","function":{"name":"edit","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\",\"old_string\":\"\\tfmt.Println(\\\"Hello, World!\\\")\",\"new_string\":\"\\tfmt.Println(\\\"Hello from crush\\\")\"}"}}]}}]} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_9d5f8a4280e149e2933f0463","index":0,"type":"function","function":{"name":"view","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\"}"}}]}}]} - data: {"id":"20251028225849bbce80c093624e4b","created":1761663529,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10765,"completion_tokens":66,"total_tokens":10831,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290038222e079ba7432e46cc","created":1761669502,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10345,"completion_tokens":124,"total_tokens":10469,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -332,15 +264,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 656.745208ms -- id: 4 + duration: 773.425958ms +- id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45472 + content_length: 43400 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_1dbca8706f7c4525b02a0169\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\",\"tool_call_id\":\"call_1dbca8706f7c4525b02a0169\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_17ae9185694a4283b7a3ae03\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_17ae9185694a4283b7a3ae03\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_3d7f707f9a9d49a19e405fa9\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\",\\\"old_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\\"new_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello from crush\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_3d7f707f9a9d49a19e405fa9\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_9d5f8a4280e149e2933f0463\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_9d5f8a4280e149e2933f0463\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -356,13 +288,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225850f47bbd31ebbc4fe0","created":1761663530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290038247c66919483de442e","created":1761669504,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225850f47bbd31ebbc4fe0","created":1761663530,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290038247c66919483de442e","created":1761669504,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225850f47bbd31ebbc4fe0","created":1761663530,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_b006b8e35fee46fea7bd6af3","index":0,"type":"function","function":{"name":"bash","arguments":"{\"command\":\"cd /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file && go run main.go\",\"description\":\"Run the updated Go program to verify the change\"}"}}]}}]} + data: {"id":"202510290038247c66919483de442e","created":1761669504,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_bddf1bd59f5f421d897b9e23","index":0,"type":"function","function":{"name":"edit","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\",\"old_string\":\"\\tfmt.Println(\\\"Hello, World!\\\")\",\"new_string\":\"\\tfmt.Println(\\\"hello from crush\\\")\"}"}}]}}]} - data: {"id":"20251028225850f47bbd31ebbc4fe0","created":1761663530,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10868,"completion_tokens":56,"total_tokens":10924,"prompt_tokens_details":{"cached_tokens":43}}} + data: {"id":"202510290038247c66919483de442e","created":1761669504,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10438,"completion_tokens":66,"total_tokens":10504,"prompt_tokens_details":{"cached_tokens":10347}}} data: [DONE] @@ -371,15 +303,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 1.152461209s -- id: 5 + duration: 841.225208ms +- id: 3 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 45948 + content_length: 43938 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_1dbca8706f7c4525b02a0169\",\"function\":{\"arguments\":\"{\\\"pattern\\\":\\\"**/main.go\\\"}\",\"name\":\"glob\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\",\"tool_call_id\":\"call_1dbca8706f7c4525b02a0169\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_17ae9185694a4283b7a3ae03\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_17ae9185694a4283b7a3ae03\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_3d7f707f9a9d49a19e405fa9\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\",\\\"old_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\\"new_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello from crush\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_3d7f707f9a9d49a19e405fa9\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_b006b8e35fee46fea7bd6af3\",\"function\":{\"arguments\":\"{\\\"command\\\":\\\"cd /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file && go run main.go\\\",\\\"description\\\":\\\"Run the updated Go program to verify the change\\\"}\",\"name\":\"bash\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"Hello from crush\\n\\n\\n/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\",\"tool_call_id\":\"call_b006b8e35fee46fea7bd6af3\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"update the main.go file by changing the print to say hello from crush\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_9d5f8a4280e149e2933f0463\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\"}\",\"name\":\"view\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\n 1|package main\\n 2|\\n 3|import \\\"fmt\\\"\\n 4|\\n 5|func main() {\\n 6|\\tfmt.Println(\\\"Hello, World!\\\")\\n 7|}\\n\\n\",\"tool_call_id\":\"call_9d5f8a4280e149e2933f0463\",\"role\":\"tool\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_bddf1bd59f5f421d897b9e23\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\\",\\\"old_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"Hello, World!\\\\\\\")\\\",\\\"new_string\\\":\\\"\\\\tfmt.Println(\\\\\\\"hello from crush\\\\\\\")\\\"}\",\"name\":\"edit\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nContent replaced in file: /tmp/crush-test/TestCoderAgent/zai-glm4.6/update_a_file/main.go\\n\\n\",\"tool_call_id\":\"call_bddf1bd59f5f421d897b9e23\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -395,47 +327,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Done"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" Updated"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" main"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".go"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" to"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" print"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" \""}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" from"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" crush"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\""}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" and"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" verified"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" it"}}]} - - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" runs"}}]} + data: {"id":"20251029003826cc70476d12494b4b","created":1761669506,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" correctly"}}]} + data: {"id":"20251029003826cc70476d12494b4b","created":1761669506,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"20251029003826cc70476d12494b4b","created":1761669506,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Done"}}]} - data: {"id":"20251028225853494586729df84ec7","created":1761663533,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10956,"completion_tokens":23,"total_tokens":10979,"prompt_tokens_details":{"cached_tokens":10707}}} + data: {"id":"20251029003826cc70476d12494b4b","created":1761669506,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10541,"completion_tokens":6,"total_tokens":10547,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -444,4 +342,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 649.239333ms + duration: 602.142042ms diff --git a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/write_tool.yaml b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/write_tool.yaml index 4f557097096a8929bf4c6a5c47c696ae8a5291f9..82842043ccd4c8f5940af9b0b844af152e857f95 100644 --- a/internal/agent/testdata/TestCoderAgent/zai-glm4.6/write_tool.yaml +++ b/internal/agent/testdata/TestCoderAgent/zai-glm4.6/write_tool.yaml @@ -24,21 +24,21 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Create"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":"Create"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" config"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" config"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":".json"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":".json"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" file"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" content"}}]} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"delta":{"role":"assistant","content":" content"}}]} - data: {"id":"2025102822594786437de495b64634","created":1761663587,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":148,"completion_tokens":10,"total_tokens":158,"prompt_tokens_details":{"cached_tokens":115}}} + data: {"id":"2025102900390532594ad0b8994aed","created":1761669545,"model":"glm-4.5-air","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":148,"completion_tokens":10,"total_tokens":158,"prompt_tokens_details":{"cached_tokens":115}}} data: [DONE] @@ -47,15 +47,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 707.152875ms + duration: 864.597458ms - id: 1 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44193 + content_length: 42976 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -71,13 +71,13 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"202510282259471da9671da55a419f","created":1761663587,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"20251029003906111b2ecffbee4364","created":1761669546,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"202510282259471da9671da55a419f","created":1761663587,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"20251029003906111b2ecffbee4364","created":1761669546,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"202510282259471da9671da55a419f","created":1761663587,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_dd1fbff62a494f31939d0025","index":0,"type":"function","function":{"name":"write","arguments":"{\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool/config.json\",\"content\":\"{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}\"}"}}]}}]} + data: {"id":"20251029003906111b2ecffbee4364","created":1761669546,"model":"glm-4.6","choices":[{"index":0,"delta":{"tool_calls":[{"id":"call_a6090481cae54835bb307c76","index":0,"type":"function","function":{"name":"write","arguments":"{\"content\":\"{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}\",\"file_path\":\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool/config.json\"}"}}]}}]} - data: {"id":"202510282259471da9671da55a419f","created":1761663587,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10641,"completion_tokens":58,"total_tokens":10699,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"20251029003906111b2ecffbee4364","created":1761669546,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"tool_calls","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10359,"completion_tokens":58,"total_tokens":10417,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -86,15 +86,15 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 705.895542ms + duration: 1.861424125s - id: 2 request: proto: HTTP/1.1 proto_major: 1 proto_minor: 1 - content_length: 44693 + content_length: 43476 host: "" - body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_dd1fbff62a494f31939d0025\",\"function\":{\"arguments\":\"{\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool/config.json\\\",\\\"content\\\":\\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\"}\",\"name\":\"write\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nFile successfully written: /tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool/config.json\\n\",\"tool_call_id\":\"call_dd1fbff62a494f31939d0025\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"web_fetch\",\"strict\":false,\"description\":\"Fetches content from a web URL (for use by sub-agents).\\n\\n\\n- Provide a URL to fetch\\n- The tool fetches the content and returns it as markdown\\n- Use this when you need to follow links from the current page\\n- After fetching, analyze the content to answer the user's question\\n\\n\\n\\n- Automatically converts HTML to markdown for easier analysis\\n- For large pages (>50KB), saves content to a temporary file and provides the path\\n- You can then use grep/view tools to search through the file\\n- Handles UTF-8 content validation\\n\\n\\n\\n- Max response size: 5MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n\\n\\n\\n- For large pages saved to files, use grep to find relevant sections first\\n- Don't fetch unnecessary pages - only when needed to answer the question\\n- Focus on extracting specific information from the fetched content\\n\\n\",\"parameters\":{\"properties\":{\"url\":{\"description\":\"The URL to fetch content from\",\"type\":\"string\"}},\"required\":[\"url\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" + body: "{\"messages\":[{\"content\":\"You are Crush, a powerful AI Assistant that runs in the CLI.\\n\\n\\nThese rules override everything else. Follow them strictly:\\n\\n1. **ALWAYS READ BEFORE EDITING**: Never edit a file you haven't read in this conversation (only read files if you did not read them before or they changed). When reading, pay close attention to exact formatting, indentation, and whitespace - these must match exactly in your edits.\\n2. **BE AUTONOMOUS**: Don't ask questions - search, read, decide, act. Complete the ENTIRE task before stopping. Never stop mid-task. Never refuse work based on scope or complexity - break it down and do it.\\n3. **TEST AFTER CHANGES**: Run tests immediately after each modification\\n4. **BE CONCISE**: Under 4 lines unless user asks for detail\\n5. **USE EXACT MATCHES**: When editing, match text exactly including whitespace, indentation, and line breaks\\n6. **NEVER COMMIT**: Unless user explicitly says \\\"commit\\\"\\n7. **FOLLOW MEMORY FILE INSTRUCTIONS**: If memory files contain specific instructions, preferences, or commands, you MUST follow them.\\n8. **NEVER ADD COMMENTS**: Only add comments if the user asked you to do so. When adding comments, focus on *why* not *what*. NEVER communicate with the user through code comments.\\n9. **SECURITY FIRST**: Only assist with defensive security tasks. Refuse to create, modify, or improve code that may be used maliciously. Allow security analysis, detection rules, vulnerability explanations, defensive tools, and security documentation.\\n10. **NO URL GUESSING**: Never generate or guess URLs unless you are confident they are for helping with programming. Only use URLs provided by the user or found in local files.\\n11. **NEVER PUSH TO REMOTE**: Don't push changes to remote repositories unless explicitly asked by the user.\\n12. **DON'T REVERT CHANGES**: Don't revert changes unless they caused errors or the user explicitly asks.\\n13. **COMPLETE THE TASK**: Never stop mid-task with \\\"Next:\\\" or \\\"Will do:\\\" statements. If you describe what needs to be done, DO IT immediately. Only stop when everything is finished.\\n14. **NEVER REFUSE BASED ON SCOPE**: Never refuse tasks because they seem large or complex. Break them into steps and complete them. Only stop if you encounter actual blocking errors (missing dependencies, compile failures, etc.), not perceived difficulty.\\n\\n\\n\\nKeep responses minimal:\\n- Under 4 lines of text (tool use doesn't count)\\n- No preamble (\\\"Here's...\\\", \\\"I'll...\\\")\\n- No postamble (\\\"Let me know...\\\", \\\"Hope this helps...\\\")\\n- One-word answers when possible\\n- No emojis ever\\n- No explanations unless user asks\\n\\nExamples:\\nuser: what is 2+2?\\nassistant: 4\\n\\nuser: list files in src/\\nassistant: [uses ls tool]\\nfoo.c, bar.c, baz.c\\n\\nuser: which file has the foo implementation?\\nassistant: src/foo.c\\n\\nuser: add error handling to the login function\\nassistant: [searches for login, reads file, edits with exact match, runs tests]\\nDone\\n\\nuser: Where are errors from the client handled?\\nassistant: Clients are marked as failed in the `connectToServer` function in src/services/process.go:712.\\n\\n\\n\\nWhen referencing specific functions or code locations, use the pattern `file_path:line_number` to help users navigate:\\n- Example: \\\"The error is handled in src/main.go:45\\\"\\n- Example: \\\"See the implementation in pkg/utils/helper.go:123-145\\\"\\n\\n\\n\\nFor every task, follow this sequence internally (don't narrate it):\\n\\n**Before acting**:\\n- Search codebase for relevant files\\n- Read files to understand current state\\n- Check memory for stored commands\\n- Identify what needs to change\\n- Use `git log` and `git blame` for additional context when needed\\n\\n**While acting**:\\n- Read entire file before editing it\\n- Before editing: verify exact whitespace and indentation from View output\\n- Use exact text for find/replace (include whitespace)\\n- Make one logical change at a time\\n- After each change: run tests\\n- If tests fail: fix immediately\\n- If edit fails: read more context, don't guess - the text must match exactly\\n- Keep going until query is completely resolved before yielding to user\\n- For longer tasks, send brief progress updates (under 10 words) BUT IMMEDIATELY CONTINUE WORKING - progress updates are not stopping points\\n\\n**Before finishing**:\\n- Verify ENTIRE query is resolved (not just first step)\\n- All described next steps must be completed\\n- Run lint/typecheck if in memory\\n- Verify all changes work\\n- Keep response under 4 lines\\n\\n**Key behaviors**:\\n- Use find_references before changing shared code\\n- Follow existing patterns (check similar files)\\n- If stuck, try different approach (don't repeat failures)\\n- Make decisions yourself (search first, don't ask)\\n- Fix problems at root cause, not surface-level patches\\n- Don't fix unrelated bugs or broken tests (mention them in final message if relevant)\\n\\n\\n\\n**Make decisions autonomously** - don't ask when you can:\\n- Search to find the answer\\n- Read files to see patterns\\n- Check similar code\\n- Infer from context\\n- Try most likely approach\\n\\n**Only stop/ask user if**:\\n- Truly ambiguous business requirement\\n- Multiple valid approaches with big tradeoffs\\n- Could cause data loss\\n- Exhausted all attempts and hit actual blocking errors\\n\\n**Never stop for**:\\n- Task seems too large (break it down)\\n- Multiple files to change (change them)\\n- Concerns about \\\"session limits\\\" (no such limits exist)\\n- Work will take many steps (do all the steps)\\n\\nExamples of autonomous decisions:\\n- File location → search for similar files\\n- Test command → check package.json/memory\\n- Code style → read existing code\\n- Library choice → check what's used\\n- Naming → follow existing names\\n\\n\\n\\n**No task is too large**:\\n- Break complex tasks into logical steps\\n- Complete each step fully before moving to next\\n- If a task has 10 parts, do all 10 parts\\n- Don't estimate effort or refuse based on scope\\n- Only stop if you hit actual errors (compile failures, missing files, etc.)\\n\\n**For large refactors or implementations**:\\n- Start with core functionality\\n- Build incrementally\\n- Test at each step\\n- Keep going until fully complete\\n\\nThere are no \\\"session limits\\\" - continue until the task is done or you hit a real blocker.\\n\\n\\n\\nCritical: ALWAYS read files before editing them in this conversation.\\n\\nWhen using edit tools:\\n1. Read the file first - note the EXACT indentation (spaces vs tabs, count)\\n2. Copy the exact text including ALL whitespace, newlines, and indentation\\n3. Include 3-5 lines of context before and after the target\\n4. Verify your old_string would appear exactly once in the file\\n5. If uncertain about whitespace, include more surrounding context\\n6. Verify edit succeeded\\n7. Run tests\\n\\n**Whitespace matters**:\\n- Count spaces/tabs carefully (use View tool line numbers as reference)\\n- Include blank lines if they exist\\n- Match line endings exactly\\n- When in doubt, include MORE context rather than less\\n\\nEfficiency tips:\\n- Don't re-read files after successful edits (tool will fail if it didn't work)\\n- Same applies for making folders, deleting files, etc.\\n\\nCommon mistakes to avoid:\\n- Editing without reading first\\n- Approximate text matches\\n- Wrong indentation (spaces vs tabs, wrong count)\\n- Missing or extra blank lines\\n- Not enough context (text appears multiple times)\\n- Trimming whitespace that exists in the original\\n- Not testing after changes\\n\\n\\n\\nThe Edit tool is extremely literal. \\\"Close enough\\\" will fail.\\n\\n**Before every edit**:\\n1. View the file and locate the exact lines to change\\n2. Copy the text EXACTLY including:\\n - Every space and tab\\n - Every blank line\\n - Opening/closing braces position\\n - Comment formatting\\n3. Include enough surrounding lines (3-5) to make it unique\\n4. Double-check indentation level matches\\n\\n**Common failures**:\\n- `func foo() {` vs `func foo(){` (space before brace)\\n- Tab vs 4 spaces vs 2 spaces\\n- Missing blank line before/after\\n- `// comment` vs `//comment` (space after //)\\n- Different number of spaces in indentation\\n\\n**If edit fails**:\\n- View the file again at the specific location\\n- Copy even more context\\n- Check for tabs vs spaces\\n- Verify line endings\\n- Try including the entire function/block if needed\\n- Never retry with guessed changes - get the exact text first\\n\\n\\n\\nWhen errors occur:\\n1. Read complete error message\\n2. Understand root cause\\n3. Try different approach (don't repeat same action)\\n4. Search for similar code that works\\n5. Make targeted fix\\n6. Test to verify\\n\\nCommon errors:\\n- Import/Module → check paths, spelling, what exists\\n- Syntax → check brackets, indentation, typos\\n- Tests fail → read test, see what it expects\\n- File not found → use ls, check exact path\\n\\n**Edit tool \\\"old_string not found\\\"**:\\n- View the file again at the target location\\n- Copy the EXACT text including all whitespace\\n- Include more surrounding context (full function if needed)\\n- Check for tabs vs spaces, extra/missing blank lines\\n- Count indentation spaces carefully\\n- Don't retry with approximate matches - get the exact text\\n\\n\\n\\nMemory files store commands, preferences, and codebase info. Update them when you discover:\\n- Build/test/lint commands\\n- Code style preferences \\n- Important codebase patterns\\n- Useful project information\\n\\n\\n\\nBefore writing code:\\n1. Check if library exists (look at imports, package.json)\\n2. Read similar code for patterns\\n3. Match existing style\\n4. Use same libraries/frameworks\\n5. Follow security best practices (never log secrets)\\n6. Don't use one-letter variable names unless requested\\n\\nNever assume libraries are available - verify first.\\n\\n**Ambition vs. precision**:\\n- New projects → be creative and ambitious with implementation\\n- Existing codebases → be surgical and precise, respect surrounding code\\n- Don't change filenames or variables unnecessarily\\n- Don't add formatters/linters/tests to codebases that don't have them\\n\\n\\n\\nAfter significant changes:\\n- Start testing as specific as possible to code changed, then broaden to build confidence\\n- Use self-verification: write unit tests, add output logs, or use debug statements to verify your solutions\\n- Run relevant test suite\\n- If tests fail, fix before continuing\\n- Check memory for test commands\\n- Run lint/typecheck if available (on precise targets when possible)\\n- For formatters: iterate max 3 times to get it right; if still failing, present correct solution and note formatting issue\\n- Suggest adding commands to memory if not found\\n- Don't fix unrelated bugs or test failures (not your responsibility)\\n\\n\\n\\n- Search before assuming\\n- Read files before editing\\n- Always use absolute paths for file operations (editing, reading, writing)\\n- Use Agent tool for complex searches\\n- Run tools in parallel when safe (no dependencies)\\n- When making multiple independent bash calls, send them in a single message with multiple tool calls for parallel execution\\n- Summarize tool output for user (they don't see it)\\n\\n\\nWhen running non-trivial bash commands (especially those that modify the system):\\n- Briefly explain what the command does and why you're running it\\n- This ensures the user understands potentially dangerous operations\\n- Simple read-only commands (ls, cat, etc.) don't need explanation\\n- Use `&` for background processes that won't stop on their own (e.g., `node server.js &`)\\n- Avoid interactive commands - use non-interactive versions (e.g., `npm init -y` not `npm init`)\\n- Combine related commands to save time (e.g., `git status && git diff HEAD && git log -n 3`)\\n\\n\\n\\n\\nBalance autonomy with user intent:\\n- When asked to do something → do it fully (including ALL follow-ups and \\\"next steps\\\")\\n- Never describe what you'll do next - just do it\\n- When asked how to approach → explain first, don't auto-implement\\n- After completing work → stop, don't explain (unless asked)\\n- Don't surprise user with unexpected actions\\n\\n\\n\\nAdapt verbosity to match the work completed:\\n\\n**Default (under 4 lines)**:\\n- Simple questions or single-file changes\\n- Casual conversation, greetings, acknowledgements\\n- One-word answers when possible\\n\\n**More detail allowed (up to 10-15 lines)**:\\n- Large multi-file changes that need walkthrough\\n- Complex refactoring where rationale adds value\\n- Tasks where understanding the approach is important\\n- When mentioning unrelated bugs/issues found\\n- Suggesting logical next steps user might want\\n\\n**What to include in verbose answers**:\\n- Brief summary of what was done and why\\n- Key files/functions changed (with `file:line` references)\\n- Any important decisions or tradeoffs made\\n- Next steps or things user should verify\\n- Issues found but not fixed\\n\\n**What to avoid**:\\n- Don't show full file contents unless explicitly asked\\n- Don't explain how to save files or copy code (user has access to your work)\\n- Don't use \\\"Here's what I did\\\" or \\\"Let me know if...\\\" style preambles/postambles\\n- Keep tone direct and factual, like handing off work to a teammate\\n\\n\\n\\nWorking directory: /tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool\\nIs directory a git repo: no\\nPlatform: linux\\nToday's date: 1/1/2025\\n\\n\\n\\n\\n\\n\\n\",\"role\":\"system\"},{\"content\":\"use write to create a new file called config.json with content '{\\\"name\\\": \\\"test\\\", \\\"version\\\": \\\"1.0.0\\\"}'\",\"role\":\"user\"},{\"content\":\"\\n\",\"tool_calls\":[{\"id\":\"call_a6090481cae54835bb307c76\",\"function\":{\"arguments\":\"{\\\"content\\\":\\\"{\\\\\\\"name\\\\\\\": \\\\\\\"test\\\\\\\", \\\\\\\"version\\\\\\\": \\\\\\\"1.0.0\\\\\\\"}\\\",\\\"file_path\\\":\\\"/tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool/config.json\\\"}\",\"name\":\"write\"},\"type\":\"function\"}],\"role\":\"assistant\"},{\"content\":\"\\nFile successfully written: /tmp/crush-test/TestCoderAgent/zai-glm4.6/write_tool/config.json\\n\",\"tool_call_id\":\"call_a6090481cae54835bb307c76\",\"role\":\"tool\"}],\"model\":\"glm-4.6\",\"max_tokens\":10000,\"stream_options\":{\"include_usage\":true},\"tool_choice\":\"auto\",\"tools\":[{\"function\":{\"name\":\"bash\",\"strict\":false,\"description\":\"Executes bash commands in persistent shell session with timeout and security measures.\\n\\n\\nUses mvdan/sh interpreter (Bash-compatible on all platforms including Windows).\\nUse forward slashes for paths: \\\"ls C:/foo/bar\\\" not \\\"ls C:\\\\foo\\\\bar\\\".\\nCommon shell builtins and core utils available on Windows.\\n\\n\\n\\n1. Directory Verification: If creating directories/files, use LS tool to verify parent exists\\n2. Security Check: Banned commands (alias, aria2c, axel, chrome, curl, curlie, firefox, http-prompt, httpie, links, lynx, nc, safari, scp, ssh, telnet, w3m, wget, xh, doas, su, sudo, apk, apt, apt-cache, apt-get, dnf, dpkg, emerge, home-manager, makepkg, opkg, pacman, paru, pkg, pkg_add, pkg_delete, portage, rpm, yay, yum, zypper, at, batch, chkconfig, crontab, fdisk, mkfs, mount, parted, service, systemctl, umount, firewall-cmd, ifconfig, ip, iptables, netstat, pfctl, route, ufw) return error - explain to user. Safe read-only commands execute without prompts\\n3. Command Execution: Execute with proper quoting, capture output\\n4. Output Processing: Truncate if exceeds 30000 characters\\n5. Return Result: Include errors, metadata with tags\\n\\n\\n\\n- Command required, timeout optional (max 600000ms/10min, default 30min if unspecified)\\n- IMPORTANT: Use Grep/Glob/Agent tools instead of 'find'/'grep'. Use View/LS tools instead of 'cat'/'head'/'tail'/'ls'\\n- Chain with ';' or '&&', avoid newlines except in quoted strings\\n- Shell state persists (env vars, virtual envs, cwd, etc.)\\n- Prefer absolute paths over 'cd' (use 'cd' only if user explicitly requests)\\n\\n\\n\\nWhen user asks to create git commit:\\n\\n1. Single message with three tool_use blocks (IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - git log (recent commit message style)\\n\\n2. Add relevant untracked files to staging. Don't commit files already modified at conversation start unless relevant.\\n\\n3. Analyze staged changes in tags:\\n - List changed/added files, summarize nature (feature/enhancement/bug fix/refactoring/test/docs)\\n - Brainstorm purpose/motivation, assess project impact, check for sensitive info\\n - Don't use tools beyond git context\\n - Draft concise (1-2 sentences) message focusing on \\\"why\\\" not \\\"what\\\"\\n - Use clear language, accurate reflection (\\\"add\\\"=new feature, \\\"update\\\"=enhancement, \\\"fix\\\"=bug fix)\\n - Avoid generic messages, review draft\\n\\n4. Create commit with Crush signature using HEREDOC:\\n git commit -m \\\"$(cat <<'EOF'\\n Commit message here.\\n\\n \U0001F498 Generated with Crush\\n\\n\\n Co-Authored-By: Crush \\n\\n EOF\\n )\\\"\\n\\n5. If pre-commit hook fails, retry ONCE. If fails again, hook preventing commit. If succeeds but files modified, MUST amend.\\n\\n6. Run git status to verify.\\n\\nNotes: Use \\\"git commit -am\\\" when possible, don't stage unrelated files, NEVER update config, don't push, no -i flags, no empty commits, return empty response.\\n\\n\\n\\nUse gh command for ALL GitHub tasks. When user asks to create PR:\\n\\n1. Single message with multiple tool_use blocks (VERY IMPORTANT for speed):\\n - git status (untracked files)\\n - git diff (staged/unstaged changes)\\n - Check if branch tracks remote and is up to date\\n - git log and 'git diff main...HEAD' (full commit history from main divergence)\\n\\n2. Create new branch if needed\\n3. Commit changes if needed\\n4. Push to remote with -u flag if needed\\n\\n5. Analyze changes in tags:\\n - List commits since diverging from main\\n - Summarize nature of changes\\n - Brainstorm purpose/motivation\\n - Assess project impact\\n - Don't use tools beyond git context\\n - Check for sensitive information\\n - Draft concise (1-2 bullet points) PR summary focusing on \\\"why\\\"\\n - Ensure summary reflects ALL changes since main divergence\\n - Clear, concise language\\n - Accurate reflection of changes and purpose\\n - Avoid generic summaries\\n - Review draft\\n\\n6. Create PR with gh pr create using HEREDOC:\\n gh pr create --title \\\"title\\\" --body \\\"$(cat <<'EOF'\\n\\n ## Summary\\n\\n <1-3 bullet points>\\n\\n ## Test plan\\n\\n [Checklist of TODOs...]\\n\\n\\n \U0001F498 Generated with Crush\\n\\n\\n EOF\\n )\\\"\\n\\nImportant:\\n\\n- Return empty response - user sees gh output\\n- Never update git config\\n\\n\\n\\nGood: pytest /foo/bar/tests\\nBad: cd /foo/bar && pytest tests\\n\\n\",\"parameters\":{\"properties\":{\"command\":{\"description\":\"The command to execute\",\"type\":\"string\"},\"description\":{\"description\":\"A brief description of what the command does\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in milliseconds (max 600000)\",\"type\":\"integer\"}},\"required\":[\"command\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"download\",\"strict\":false,\"description\":\"Downloads binary data from URL and saves to local file.\\n\\n\\n- Provide URL to download from\\n- Specify local file path where content should be saved\\n- Optional timeout for request\\n\\n\\n\\n- Downloads any file type (binary or text)\\n- Auto-creates parent directories if missing\\n- Handles large files efficiently with streaming\\n- Sets reasonable timeouts to prevent hanging\\n- Validates input parameters before requests\\n\\n\\n\\n- Max file size: 100MB\\n- Only supports HTTP and HTTPS protocols\\n- Cannot handle authentication or cookies\\n- Some websites may block automated requests\\n- Will overwrite existing files without warning\\n\\n\\n\\n- Use absolute paths or paths relative to working directory\\n- Set appropriate timeouts for large files or slow connections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The local file path where the downloaded content should be saved\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 600)\",\"type\":\"integer\"},\"url\":{\"description\":\"The URL to download from\",\"type\":\"string\"}},\"required\":[\"url\",\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"edit\",\"strict\":false,\"description\":\"Edits files by replacing text, creating new files, or deleting content. For moving/renaming use Bash 'mv'. For large edits use Write tool.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. For new files: Use LS tool to verify parent directory exists\\n3. **CRITICAL**: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. old_string: Text to replace (must match exactly including whitespace/indentation)\\n3. new_string: Replacement text\\n4. replace_all: Replace all occurrences (default false)\\n\\n\\n\\n\\n- Create file: provide file_path + new_string, leave old_string empty\\n- Delete content: provide file_path + old_string, leave new_string empty\\n \\n\\n\\nEXACT MATCHING: The tool is extremely literal. Text must match **EXACTLY**\\n\\n- Every space and tab character\\n- Every blank line\\n- Every newline character\\n- Indentation level (count the spaces/tabs)\\n- Comment spacing (`// comment` vs `//comment`)\\n- Brace positioning (`func() {` vs `func(){`)\\n\\nCommon failures:\\n\\n```\\nExpected: \\\" func foo() {\\\" (4 spaces)\\nProvided: \\\" func foo() {\\\" (2 spaces) ❌ FAILS\\n\\nExpected: \\\"}\\\\n\\\\nfunc bar() {\\\" (2 newlines)\\nProvided: \\\"}\\\\nfunc bar() {\\\" (1 newline) ❌ FAILS\\n\\nExpected: \\\"// Comment\\\" (space after //)\\nProvided: \\\"//Comment\\\" (no space) ❌ FAILS\\n```\\n\\nUNIQUENESS (when replace_all=false): old_string MUST uniquely identify target instance\\n\\n- Include 3-5 lines context BEFORE and AFTER change point\\n- Include exact whitespace, indentation, surrounding code\\n- If text appears multiple times, add more context to make it unique\\n\\nSINGLE INSTANCE: Tool changes ONE instance when replace_all=false\\n\\n- For multiple instances: set replace_all=true OR make separate calls with unique context\\n- Plan calls carefully to avoid conflicts\\n\\nVERIFICATION BEFORE USING: Before every edit\\n\\n1. View the file and locate exact target location\\n2. Check how many instances of target text exist\\n3. Copy the EXACT text including all whitespace\\n4. Verify you have enough context for unique identification\\n5. Double-check indentation matches (count spaces/tabs)\\n6. Plan separate calls or use replace_all for multiple changes\\n \\n\\n\\nTool fails if:\\n- old_string matches multiple locations and replace_all=false\\n- old_string doesn't match exactly (including whitespace)\\n- Insufficient context causes wrong instance change\\n- Indentation is off by even one space\\n- Missing or extra blank lines\\n- Wrong tabs vs spaces\\n\\n\\n\\nIf you get \\\"old_string not found in file\\\":\\n\\n1. **View the file again** at the specific location\\n2. **Copy more context** - include entire function if needed\\n3. **Check whitespace**:\\n - Count indentation spaces/tabs\\n - Look for blank lines\\n - Check for trailing spaces\\n4. **Verify character-by-character** that your old_string matches\\n5. **Never guess** - always View the file to get exact text\\n \\n\\n\\n\\n- Ensure edits result in correct, idiomatic code\\n- Don't leave code in broken state\\n- Use absolute file paths (starting with /)\\n- Use forward slashes (/) for cross-platform compatibility\\n- Multiple edits to same file: send all in single message with multiple tool calls\\n- **When in doubt, include MORE context rather than less**\\n- Match the existing code style exactly (spaces, tabs, blank lines)\\n \\n\\n\\nBefore submitting an edit, verify:\\n\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if they exist\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3-5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all)\\n- [ ] Copied text character-for-character, not approximated\\n \\n\\n\\n✅ Correct: Exact match with context\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n\\nnew_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n // New validation\\\\n if len(input) > 1000 {\\\\n return errors.New(\\\\\\\"input too long\\\\\\\")\\\\n }\\\\n return nil\\\\n}\\\"\\n```\\n\\n❌ Incorrect: Not enough context\\n\\n```\\nold_string: \\\"return nil\\\" // Appears many times!\\n```\\n\\n❌ Incorrect: Wrong indentation\\n\\n```\\nold_string: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 2 spaces\\n// But file actually has: \\\" if input == \\\\\\\"\\\\\\\" {\\\" // 4 spaces\\n```\\n\\n✅ Correct: Including context to make unique\\n\\n```\\nold_string: \\\"func ProcessData(input string) error {\\\\n if input == \\\\\\\"\\\\\\\" {\\\\n return errors.New(\\\\\\\"empty input\\\\\\\")\\\\n }\\\\n return nil\\\"\\n```\\n\\n\\n\\n\\n\\n- Forward slashes work throughout (C:/path/file)\\n- File permissions handled automatically\\n- Line endings converted automatically (\\\\n ↔ \\\\r\\\\n)\\n \\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"},\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false)\",\"type\":\"boolean\"}},\"required\":[\"file_path\",\"old_string\",\"new_string\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"multiedit\",\"strict\":false,\"description\":\"Makes multiple edits to a single file in one operation. Built on Edit tool for efficient multiple find-and-replace operations. Prefer over Edit tool for multiple edits to same file.\\n\\n\\n1. Use View tool to understand file contents and context\\n2. Verify directory path is correct\\n3. CRITICAL: Note exact whitespace, indentation, and formatting from View output\\n\\n\\n\\n1. file_path: Absolute path to file (required)\\n2. edits: Array of edit operations, each containing:\\n - old_string: Text to replace (must match exactly including whitespace/indentation)\\n - new_string: Replacement text\\n - replace_all: Replace all occurrences (optional, defaults to false)\\n\\n\\n\\n- Edits applied sequentially in provided order.\\n- Each edit operates on result of previous edit.\\n- ATOMIC: If any single edit fails, the entire operation fails and no changes are applied.\\n- Ideal for several changes to different parts of same file.\\n\\n\\n\\nAll instructions from the Edit tool documentation apply verbatim to every edit item:\\n- Critical requirements for exact matching and uniqueness\\n- Warnings and common failures (tabs vs spaces, blank lines, brace placement, etc.)\\n- Verification steps before using, recovery steps, best practices, and whitespace checklist\\nUse the same level of precision as Edit. Multiedit often fails due to formatting mismatches—double-check whitespace for every edit.\\n\\n\\n\\n1. Apply Edit tool rules to EACH edit (see edit.md).\\n2. Edits are atomic—either all succeed or none are applied.\\n3. Plan sequence carefully: earlier edits change the file content that later edits must match.\\n4. Ensure each old_string is unique at its application time (after prior edits).\\n\\n\\n\\n1. View the file and copy exact text (including whitespace) for each target.\\n2. Check how many instances each old_string has BEFORE the sequence starts.\\n3. Dry-run mentally: after applying edit #N, will edit #N+1 still match? Adjust old_string/new_string accordingly.\\n4. Prefer fewer, larger context blocks over many tiny fragments that are easy to misalign.\\n5. If edits are independent, consider separate multiedit batches per logical region.\\n\\n\\n\\n- Operation fails if any old_string doesn’t match exactly (including whitespace) or equals new_string.\\n- Earlier edits can invalidate later matches (added/removed spaces, lines, or reordered text).\\n- Mixed tabs/spaces, trailing spaces, or missing blank lines commonly cause failures.\\n- replace_all may affect unintended regions—use carefully or provide more context.\\n\\n\\n\\nIf the operation fails:\\n1. Identify the first failing edit (start from top; test subsets to isolate).\\n2. View the file again and copy more surrounding context for that edit.\\n3. Recalculate later old_string values based on the file state AFTER preceding edits.\\n4. Reduce the batch (apply earlier stable edits first), then follow up with the rest.\\n\\n\\n\\n- Ensure all edits result in correct, idiomatic code; don’t leave code broken.\\n- Use absolute file paths (starting with /).\\n- Use replace_all only when you’re certain; otherwise provide unique context.\\n- Match existing style exactly (spaces, tabs, blank lines).\\n- Test after the operation; if it fails, fix and retry in smaller chunks.\\n\\n\\n\\nFor EACH edit, verify:\\n- [ ] Viewed the file first\\n- [ ] Counted indentation spaces/tabs\\n- [ ] Included blank lines if present\\n- [ ] Matched brace/bracket positioning\\n- [ ] Included 3–5 lines of surrounding context\\n- [ ] Verified text appears exactly once (or using replace_all deliberately)\\n- [ ] Copied text character-for-character, not approximated\\n\\n\\n\\n✅ Correct: Sequential edits where the second match accounts for the first change\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n doNew()\\\\n}\\\",\\n },\\n {\\n // Uses context that still exists AFTER the first replacement\\n old_string: \\\"func B() {\\\\n callA()\\\\n}\\\",\\n new_string: \\\"func B() {\\\\n callA()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n❌ Incorrect: Second old_string no longer matches due to whitespace change introduced by the first edit\\n\\n```\\nedits: [\\n {\\n old_string: \\\"func A() {\\\\n doOld()\\\\n}\\\",\\n new_string: \\\"func A() {\\\\n\\\\n doNew()\\\\n}\\\", // Added extra blank line\\n },\\n {\\n old_string: \\\"func A() {\\\\n doNew()\\\\n}\\\", // Missing the new blank line, will FAIL\\n new_string: \\\"func A() {\\\\n doNew()\\\\n logChange()\\\\n}\\\",\\n },\\n]\\n```\\n\\n\",\"parameters\":{\"properties\":{\"edits\":{\"description\":\"Array of edit operations to perform sequentially on the file\",\"items\":{\"new_string\":{\"description\":\"The text to replace it with\",\"type\":\"string\"},\"old_string\":{\"description\":\"The text to replace\",\"type\":\"string\"},\"replace_all\":{\"description\":\"Replace all occurrences of old_string (default false).\",\"type\":\"boolean\"}},\"type\":\"array\"},\"file_path\":{\"description\":\"The absolute path to the file to modify\",\"type\":\"string\"}},\"required\":[\"file_path\",\"edits\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"glob\",\"strict\":false,\"description\":\"Fast file pattern matching tool that finds files by name/pattern, returning paths sorted by modification time (newest first).\\n\\n\\n- Provide glob pattern to match against file paths\\n- Optional starting directory (defaults to current working directory)\\n- Results sorted with most recently modified files first\\n\\n\\n\\n- '\\\\*' matches any sequence of non-separator characters\\n- '\\\\*\\\\*' matches any sequence including separators\\n- '?' matches any single non-separator character\\n- '[...]' matches any character in brackets\\n- '[!...]' matches any character not in brackets\\n\\n\\n\\n- '*.js' - JavaScript files in current directory\\n- '**/*.js' - JavaScript files in any subdirectory\\n- 'src/**/*.{ts,tsx}' - TypeScript files in src directory\\n- '*.{html,css,js}' - HTML, CSS, and JS files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Does not search file contents (use Grep for that)\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Path separators handled automatically (/ and \\\\ work)\\n- Uses ripgrep (rg) if available, otherwise Go implementation\\n- Patterns should use forward slashes (/) for compatibility\\n\\n\\n\\n- Combine with Grep: find files with Glob, search contents with Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine pattern if needed\\n\\n\",\"parameters\":{\"properties\":{\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The glob pattern to match files against\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"grep\",\"strict\":false,\"description\":\"Fast content search tool that finds files containing specific text/patterns, returning matching paths sorted by modification time (newest first).\\n\\n\\n- Provide regex pattern to search within file contents\\n- Set literal_text=true for exact text with special characters (recommended for non-regex users)\\n- Optional starting directory (defaults to current working directory)\\n- Optional include pattern to filter which files to search\\n- Results sorted with most recently modified files first\\n\\n\\n\\nWhen literal_text=false (supports standard regex):\\n\\n- 'function' searches for literal text \\\"function\\\"\\n- 'log\\\\..\\\\*Error' finds text starting with \\\"log.\\\" and ending with \\\"Error\\\"\\n- 'import\\\\s+.\\\\*\\\\s+from' finds import statements in JavaScript/TypeScript\\n\\n\\n\\n- '\\\\*.js' - Only search JavaScript files\\n- '\\\\*.{ts,tsx}' - Only search TypeScript files\\n- '\\\\*.go' - Only search Go files\\n\\n\\n\\n- Results limited to 100 files (newest first)\\n- Performance depends on number of files searched\\n- Very large binary files may be skipped\\n- Hidden files (starting with '.') skipped\\n\\n\\n\\n- Respects .gitignore patterns to skip ignored files/directories\\n- Respects .crushignore patterns for additional ignore rules\\n- Both ignore files auto-detected in search root directory\\n\\n\\n\\n- Uses ripgrep (rg) if available for better performance\\n- Falls back to Go implementation if ripgrep unavailable\\n- File paths normalized automatically for compatibility\\n\\n\\n\\n- For faster searches: use Glob to find relevant files first, then Grep\\n- For iterative exploration requiring multiple searches, consider Agent tool\\n- Check if results truncated and refine search pattern if needed\\n- Use literal_text=true for exact text with special characters (dots, parentheses, etc.)\\n\\n\",\"parameters\":{\"properties\":{\"include\":{\"description\":\"File pattern to include in the search (e.g. \\\"*.js\\\", \\\"*.{ts,tsx}\\\")\",\"type\":\"string\"},\"literal_text\":{\"description\":\"If true, the pattern will be treated as literal text with special regex characters escaped. Default is false.\",\"type\":\"boolean\"},\"path\":{\"description\":\"The directory to search in. Defaults to the current working directory.\",\"type\":\"string\"},\"pattern\":{\"description\":\"The regex pattern to search for in file contents\",\"type\":\"string\"}},\"required\":[\"pattern\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"ls\",\"strict\":false,\"description\":\"Shows files and subdirectories in tree structure for exploring project organization.\\n\\n\\n- Provide path to list (defaults to current working directory)\\n- Optional glob patterns to ignore\\n- Results displayed in tree structure\\n\\n\\n\\n- Hierarchical view of files and directories\\n- Auto-skips hidden files/directories (starting with '.')\\n- Skips common system directories like __pycache__\\n- Can filter files matching specific patterns\\n\\n\\n\\n- Results limited to 1000 files\\n- Large directories truncated\\n- No file sizes or permissions shown\\n- Cannot recursively list all directories in large projects\\n\\n\\n\\n- Hidden file detection uses Unix convention (files starting with '.')\\n- Windows hidden files (with hidden attribute) not auto-skipped\\n- Common Windows directories (System32, Program Files) not in default ignore\\n- Path separators handled automatically (/ and \\\\ work)\\n\\n\\n\\n- Use Glob for finding files by name patterns instead of browsing\\n- Use Grep for searching file contents\\n- Combine with other tools for effective exploration\\n\\n\",\"parameters\":{\"properties\":{\"depth\":{\"description\":\"The maximum depth to traverse\",\"type\":\"integer\"},\"ignore\":{\"description\":\"List of glob patterns to ignore\",\"items\":{},\"type\":\"array\"},\"path\":{\"description\":\"The path to the directory to list (defaults to current working directory)\",\"type\":\"string\"}},\"required\":[],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"sourcegraph\",\"strict\":false,\"description\":\"Search code across public repositories using Sourcegraph's GraphQL API.\\n\\n\\n- Provide search query using Sourcegraph syntax\\n- Optional result count (default: 10, max: 20)\\n- Optional timeout for request\\n\\n\\n\\n- \\\"fmt.Println\\\" - exact matches\\n- \\\"file:.go fmt.Println\\\" - limit to Go files\\n- \\\"repo:^github\\\\.com/golang/go$ fmt.Println\\\" - specific repos\\n- \\\"lang:go fmt.Println\\\" - limit to Go code\\n- \\\"fmt.Println AND log.Fatal\\\" - combined terms\\n- \\\"fmt\\\\.(Print|Printf|Println)\\\" - regex patterns\\n- \\\"\\\\\\\"exact phrase\\\\\\\"\\\" - exact phrase matching\\n- \\\"-file:test\\\" or \\\"-repo:forks\\\" - exclude matches\\n\\n\\n\\nRepository: repo:name, repo:^exact$, repo:org/repo@branch, -repo:exclude, fork:yes, archived:yes, visibility:public\\nFile: file:\\\\.js$, file:internal/, -file:test, file:has.content(text)\\nContent: content:\\\"exact\\\", -content:\\\"unwanted\\\", case:yes\\nType: type:symbol, type:file, type:path, type:diff, type:commit\\nTime: after:\\\"1 month ago\\\", before:\\\"2023-01-01\\\", author:name, message:\\\"fix\\\"\\nResult: select:repo, select:file, select:content, count:100, timeout:30s\\n\\n\\n\\n- \\\"file:.go context.WithTimeout\\\" - Go code using context.WithTimeout\\n- \\\"lang:typescript useState type:symbol\\\" - TypeScript React useState hooks\\n- \\\"repo:^github\\\\.com/kubernetes/kubernetes$ pod list type:file\\\" - Kubernetes pod files\\n- \\\"file:Dockerfile (alpine OR ubuntu) -content:alpine:latest\\\" - Dockerfiles with base images\\n\\n\\n\\n- \\\"term1 AND term2\\\" - both terms\\n- \\\"term1 OR term2\\\" - either term\\n- \\\"term1 NOT term2\\\" - term1 but not term2\\n- \\\"term1 and (term2 or term3)\\\" - grouping with parentheses\\n\\n\\n\\n- Only searches public repositories\\n- Rate limits may apply\\n- Complex queries take longer\\n- Max 20 results per query\\n\\n\\n\\n- Use specific file extensions to narrow results\\n- Add repo: filters for targeted searches\\n- Use type:symbol for function/method definitions\\n- Use type:file to find relevant files\\n\\n\",\"parameters\":{\"properties\":{\"context_window\":{\"description\":\"The context around the match to return (default: 10 lines)\",\"type\":\"integer\"},\"count\":{\"description\":\"Optional number of results to return (default: 10, max: 20)\",\"type\":\"integer\"},\"query\":{\"description\":\"The Sourcegraph search query\",\"type\":\"string\"},\"timeout\":{\"description\":\"Optional timeout in seconds (max 120)\",\"type\":\"integer\"}},\"required\":[\"query\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"view\",\"strict\":false,\"description\":\"Reads and displays file contents with line numbers for examining code, logs, or text data.\\n\\n\\n- Provide file path to read\\n- Optional offset: start reading from specific line (0-based)\\n- Optional limit: control lines read (default 2000)\\n- Don't use for directories (use LS tool instead)\\n\\n\\n\\n- Displays contents with line numbers\\n- Can read from any file position using offset\\n- Handles large files by limiting lines read\\n- Auto-truncates very long lines for display\\n- Suggests similar filenames when file not found\\n\\n\\n\\n- Max file size: 250KB\\n- Default limit: 2000 lines\\n- Lines >2000 chars truncated\\n- Cannot display binary files/images (identifies them)\\n\\n\\n\\n- Handles Windows (CRLF) and Unix (LF) line endings\\n- Works with forward slashes (/) and backslashes (\\\\)\\n- Auto-detects text encoding for common formats\\n\\n\\n\\n- Use with Glob to find files first\\n- For code exploration: Grep to find relevant files, then View to examine\\n- For large files: use offset parameter for specific sections\\n\\n\",\"parameters\":{\"properties\":{\"file_path\":{\"description\":\"The path to the file to read\",\"type\":\"string\"},\"limit\":{\"description\":\"The number of lines to read (defaults to 2000)\",\"type\":\"integer\"},\"offset\":{\"description\":\"The line number to start reading from (0-based)\",\"type\":\"integer\"}},\"required\":[\"file_path\"],\"type\":\"object\"}},\"type\":\"function\"},{\"function\":{\"name\":\"write\",\"strict\":false,\"description\":\"Creates or updates files in filesystem for saving/modifying text content.\\n\\n\\n- Provide file path to write\\n- Include content to write to file\\n- Tool creates necessary parent directories automatically\\n\\n\\n\\n- Creates new files or overwrites existing ones\\n- Auto-creates parent directories if missing\\n- Checks if file modified since last read for safety\\n- Avoids unnecessary writes when content unchanged\\n\\n\\n\\n- Read file before writing to avoid conflicts\\n- Cannot append (rewrites entire file)\\n\\n\\n\\n- Use forward slashes (/) for compatibility\\n\\n\\n\\n- Use View tool first to examine existing files before modifying\\n- Use LS tool to verify location when creating new files\\n- Combine with Glob/Grep to find and modify multiple files\\n- Include descriptive comments when changing existing code\\n\\n\",\"parameters\":{\"properties\":{\"content\":{\"description\":\"The content to write to the file\",\"type\":\"string\"},\"file_path\":{\"description\":\"The path to the file to write\",\"type\":\"string\"}},\"required\":[\"file_path\",\"content\"],\"type\":\"object\"}},\"type\":\"function\"}],\"stream\":true}" headers: Accept: - application/json @@ -110,19 +110,25 @@ interactions: proto_minor: 0 content_length: -1 body: |+ - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"\n"}}]} - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"\n"}}]} - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"File"}}]} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"Created"}}]} - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" created"}}]} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" config"}}]} - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" successfully"}}]} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":".json"}}]} - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" with"}}]} - data: {"id":"20251028225949e856bf8ac131426b","created":1761663589,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10734,"completion_tokens":9,"total_tokens":10743,"prompt_tokens_details":{"cached_tokens":10579}}} + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" specified"}}]} + + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":" content"}}]} + + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"delta":{"role":"assistant","content":"."}}]} + + data: {"id":"202510290039077c46cbc33a734487","created":1761669547,"model":"glm-4.6","choices":[{"index":0,"finish_reason":"stop","delta":{"role":"assistant","content":""}}],"usage":{"prompt_tokens":10452,"completion_tokens":12,"total_tokens":10464,"prompt_tokens_details":{"cached_tokens":10297}}} data: [DONE] @@ -131,4 +137,4 @@ interactions: - text/event-stream;charset=UTF-8 status: 200 OK code: 200 - duration: 611.143584ms + duration: 540.991042ms